Analysis of Variable Pitch Air Turbines For Oscillating Water Col PDF

University of Wollongong
Research Online
University of Wollongong Thesis Collection University of Wollongong Thesis Collections
2011
Analysis of variable pitch air turbines for oscillating

water column (OWC) wave energy converters
Andrei Gareev
University of Wollongong
Recommended Citation
Gareev, Andrei, Analysis of variable pitch air turbines for oscillating water column (OWC) wave energy converters, Doctor of
Philosophy thesis, University of Wollongong. School of Mechanical, Materials and Mechatronic Engineering, University of
Wollongong, 2011. http://ro.uow.edu.au/theses/3418
Research Online is the open access institutional repository for the

University of Wollongong. For further information contact Manager
Repository Services: morgan@uow.edu.au.
ANALYSIS OF VARIABLE PITCH AIR TURBINES FOR
OSCILLATING WATER COLUMN (OWC) WAVE ENERGY
CONVERTERS
A thesis submitted in partial fulfilment of the requirements

for the award of the degree
DOCTOR OF PHILOSOPHY
from
UNIVERSITY OF WOLLONGONG
by
ANDREI GAREEV
SCHOOL OF MECHANICAL, MATERIALS AND MECHATRONIC

ENGINEERING
2011
Declaration
DECLARATION
I, Andrei Gareev, declare that this thesis, submitted in fulfilment of the requirements for
the award of Doctor of Philosophy, in the School Mechanical, Materials and Mechatronic
Engineering, University of Wollongong, is wholly my own work unless otherwise
referenced or acknowledged. The thesis has not been submitted for qualifications at any
other academic institution.
Andrei Gareev
“__” ______ 2011
i
Abstract
ABSTRACT
The work described in this thesis was focused on computational fluid dynamic (CFD)
modelling of a range of axial flow air turbines designed to service Oscillating Water
Column (OWC) wave energy converters. In particular, the research concentrated on: a)
analysis of the performance of turbines such as the Denniss-Auld turbine with variable
pitch blades; b) the influence of the design of components such as the nozzle-diffuser, the
nacelle and the OWC chamber on turbines coupled to OWC. This work was primarily
motivated by the fact that very little research and data had been previously reported on
these issues and that much work has yet to be done to develop design methods to
optimise such systems by increasing their efficiency and extending their operational
range and technical reliability.
Preliminary investigations into simplified turbine analysis via the Blade Element
Momentum (BEM) method revealed a number of “cascade” issues having particular
importance for analysis of axial flow turbines, which had not been previously examined
with the aid of CFD. Thus, one of the significant outcomes of the present study was that
CFD analysis confirmed the applicability of Weinig’s analytical and inviscid prediction
of the cascade lift interference factor, k0, for linear cascades of practical aerofoils (e.g.
NACA0012 and NACA0021) providing angles of attack are less than about 10°.
Weinig’s theory holds that the cascade lift interference factor, k0, for an infinite cascade
is independent of the angle of attack, α, and is only a function of the stagger angle, γ, and
ratio of s/c (inverse of solidity). However, through the present CFD study it was found
that the cascade interference factor does indeed depend on the angle of incidence due to
practical issues such as finite aerofoil thickness, stall, drag, etc, which could not be
accounted for in Weinig’s inviscid formulation. Thus, Weinig’s inviscid model must be
ii
Abstract
used with caution to predict the lift of blades in a cascade at angles of incidence, αm >
10º.
A second important outcome of the present study was the development of a new non-
dimensional formulation of blade element analysis which can be used to predict the
performance of OWC axial flow turbines using only generic/non-dimensional input
parameters such as hub-to-tip ratio and non-dimensional axial velocity or flow factor, φ.
Other results from the CFD study were that the efficiency of the Denniss-Auld turbine is
strongly influenced by variation of design parameters such as tip clearance, tc, hub-to-tip
ratio, h, and number of blades, Ν. Adjustment of these design parameters so that
tc = 2.3mm, h = 0.62, and N = 13 resulted in a predicted increase in turbine efficiency by
up 5.5% compared to the actual/baseline rotor configuration. CFD modelling to ascertain
the influence of the non-dimensional thickness of the blades in a Denniss-Auld turbine
demonstrated that thinner blades produce higher turbine efficiencies for large blade
stagger angles. Moreover, in general thinner blades will result in higher maximum turbine
efficiencies for a given stagger angle. This was in contrast to previous research on Well’s
turbines.
Three dimensional CFD modelling was carried out on a full scale OWC wave energy
converter case study, which was based on a demonstration plant built and commissioned
by Oceanlinx at Port Kembla, Australia, in 2006. This case study demonstrated that the
CFD technique can be successfully applied to the analysis and optimisation of the major
components of a turbine/OWC system, such as the nozzle-diffuser, nacelle and rotor. It
was found that the pneumatic efficiency, ηpneu, of the case study system depended on the
volume flow rate through the system. Other issues investigated include the influence of
the 90° turn that the air flow must make between the OWC chamber and the inlet to the
turbine for OWC systems employing a turbine with a horizontal axis of rotation. CFD
iii
Abstract
analysis demonstrated that the horizontal air ducts with converging nozzle section
appeared to have no detrimental effect on the uniformity of the axial velocity as a
function of circumferential position around the turbine.
Analysis of the experimental data obtained during sea-trial tests of the full scale
Denniss-Auld turbine in 2006 revealed that the most efficient turbine operation was
achieved with blades staggered at an angle of approximately 60° relative to the plane of
the rotor. It was also found that the 3D CFD analysis improved the correlation between
numerical results and experimental efficiency data for the full scale Dennis-Auld turbine
by approximately 11% as compared to the BEM method.
iv
Acknowledgements
ACKNOWLEDGEMENTS
Among the many people who helped me during the time of my candidature, I would wish
to acknowledge the assistance, guidance and support of the following people:
• Professor Paul Cooper, who has been my supervisor since I enrolled for a PhD. I
would like to thank him for his collaboration in the development of a non-
dimensional blade element model as well as in all written papers. I also wish to
thank him for all data and valuable information provided. In particular I would
like to thank him for comments on the present structure of the thesis.
• Doctor Buyung Kasasih who became my co-supervisor in 2007 and gave me great
assistance in applying CFD techniques for the purpose of the present project. I
also would like to thank him for providing constructive criticism regarding the
drafts.
• Doctor Brad Stappenbelt, who gave me valuable advice and encouragement
during my final year.
• Oceanlinx for provision of experimental data on the performance of their full
scale turbine and OWC system at Port Kembla.
• My wife Natalia for her patience and support throughout all years of my
candidature.
• My Mum for her help with my son Pasha who was born in 2007 and required
much attention.
• Mr Patrick McGowan who listened and encouraged me throughout my project.
v
Table of Contents
TABLE OF CONTENTS
DECLARATION ........................................................................................................................................... I
ABSTRACT ................................................................................................................................................. II
ACKNOWLEDGEMENTS .........................................................................................................................V
TABLE OF CONTENTS ........................................................................................................................... VI
TABLE OF FIGURES ............................................................................................................................... XI
LIST OF TABLES..................................................................................................................................XXV
NOMENCLATURE ............................................................................................................................. XXVI
CHAPTER 1 - INTRODUCTION........................................................................................................1
1.1 MOTIVATION AND OBJECTIVES ......................................................................................................3

1.1.1 Identification of research issues ..............................................................................................3
1.1.2 Primary objectives ...................................................................................................................5
CHAPTER 2 - DEVELOPMENT OF OWC TURBINES..................................................................7
2.1 OWC TECHNOLOGY AS A PRACTICAL MEANS OF WAVE ENERGY CONVERSION ..............................7

2.1.1 Classification of OWC wave energy converters ......................................................................7
2.1.2 Working principles of OWC wave energy devices ...................................................................8
2.1.3 Technological features of OWCs ...........................................................................................10
2.1.4 Design variations of OWCs ...................................................................................................17
2.1.4.1 Shore-mounted OWC systems ................................................................................................... 17
2.1.4.2 Sea-bed mounted OWC devices................................................................................................. 19
2.1.4.3 Floating OWC plants ................................................................................................................. 21
2.2 DESIGN REQUIREMENTS FOR OPTIMAL OPERATION OF OWC SYSTEMS ........................................24
2.2.1 Modelling of the OWC system ...............................................................................................25
2.2.1.1 Coupling of the hydrodynamic and aerodynamic domains ........................................................ 25
2.2.1.2 Parameters influencing performance of the OWC...................................................................... 28
2.2.1.3 Parameters influencing applied damping of OWC turbines ....................................................... 30
2.3 DEVELOPMENT OF AIR TURBINES FOR OWC APPLICATIONS ........................................................32
2.3.1 Air turbines with fixed pitch blades .......................................................................................33
2.3.1.1 Monoplane Wells turbines ......................................................................................................... 33
2.3.1.2 Key design parameters affecting the Wells turbine performance............................................... 35
2.3.1.3 The biplane Wells turbine .......................................................................................................... 43
2.3.1.4 Fixed guide vane Wells turbine monoplane ............................................................................... 45
2.3.1.5 The contra-rotating Wells turbine .............................................................................................. 46
vi
Table of Contents
2.3.1.6 Air turbines with fixed non-zero stagger blade angles ............................................................... 48
2.3.2 Air turbines with variable pitch angle blades........................................................................50
2.3.2.1 The Wells turbine with self-pitch controlled blades................................................................... 50
2.3.2.2 The Wells turbine with variable pitch angle blades.................................................................... 51
2.3.2.3 The Denniss-Auld turbine with variable rotor geometry............................................................ 53
2.3.3 The impulse turbine ...............................................................................................................56
2.3.3.1 Variable radius turbines ............................................................................................................. 60
2.3.4 Major OWC turbines tested in real sea conditions................................................................61
2.3.5 Design methodologies: traditional and contemporary ..........................................................62
2.3.5.1 Free vortex radial equilibrium theory......................................................................................... 62
2.3.5.2 Blade element momentum theory............................................................................................... 64
2.4 CFD MODELLING OF OWC TURBINES .........................................................................................65
2.4.1 Historical overview................................................................................................................65
2.4.2 Contemporary status..............................................................................................................68
2.5 CONCLUDING REMARKS ..............................................................................................................72
CHAPTER 3 - METHODOLOGIES FOR AERODYNAMIC ANALYSIS OF OWC AXIAL

FLOW TURBINES .....................................................................................................................................74
3.1 BLADE ELEMENT MOMENTUM (BEM) ANALYSIS.........................................................................74

3.1.1 Dimensional characterisation of OWC turbines....................................................................77
3.2 CFD ANALYSIS OF OWC TURBINES ............................................................................................78
3.2.1 ANSYS CFX ...........................................................................................................................78
3.2.2 CFD methodology adopted in the present study....................................................................80
3.2.3 CFD governing equations......................................................................................................80
3.2.4 Domain discretization methods .............................................................................................81
3.2.4.1 Discretization errors ................................................................................................................... 81
3.2.5 Turbulence modelling ............................................................................................................82
3.2.5.1 Reynolds Averaged Navier-Stokes (RANS) equations .............................................................. 82
3.2.5.2 The k-ε turbulence model ........................................................................................................... 84
3.2.5.3 Alternative turbulence models ................................................................................................... 85
3.2.5.4 Near-wall treatments .................................................................................................................. 87
3.2.6 Modelling of CFD geometry ..................................................................................................88
3.2.7 Mesh generation ....................................................................................................................89
3.2.7.1 Types of grids ............................................................................................................................ 89
3.2.7.2 Control features of the meshing process .................................................................................... 92
3.2.7.3 Grid independence ..................................................................................................................... 93
3.2.8 Selection of physics and fluid properties ...............................................................................94
3.2.9 Modelling rotors and rotating machinery..............................................................................94
3.2.10 Running the Solver............................................................................................................96
3.2.11 Verification and validation of CFD results.......................................................................97
3.2.12 Post-processing of turbomachinery ..................................................................................98
3.3 DESIGN PROCESS OVERVIEW FOR MODERN OWC TURBINES .......................................................98
vii
Table of Contents
CHAPTER 4 - CFD ANALYSIS OF ISOLATED AEROFOILS ..................................................100
4.1 EXPERIMENTAL CL AND CD COEFFICIENTS FOR VARIOUS AEROFOIL PROFILES...........................100

4.2 CFD MODELLING OF AERODYNAMIC COEFFICIENTS FOR ISOLATED AEROFOILS.........................106
4.2.1 Calculation procedure for CL and CD coefficients...............................................................106
4.2.2 Aerofoil NACA 64A010........................................................................................................106
4.2.3 Aerofoil NACA 6409 ............................................................................................................111
4.2.4 Aerofoil NACA 0012 ............................................................................................................113
4.3 CFD MODELLING OF AERODYNAMIC COEFFICIENTS NECESSARY FOR THE PRESENT STUDY .......116
4.3.1 Aerofoil for the1/3rd scale Denniss-Auld turbine .................................................................116
4.3.2 Aerofoil for the full scale Denniss-Auld turbine ..................................................................121
4.4 CONCLUDING REMARKS ............................................................................................................127
CHAPTER 5 - TWO DIMENSIONAL LINEAR CASCADES AND DETERMINATION OF

INTERFERENCE FACTORS .................................................................................................................128
5.1 ANALYTICAL PREDICTION OF CASCADE INTERFERENCE ............................................................130

5.2 CFD MODELLING OF CASCADES WITH STAGGERED BLADES ......................................................136
5.2.1 Cascades with aerofoils staggered at 0º ..............................................................................136
5.2.2 Aerofoils in a cascades with non-zero stagger angle...........................................................147
5.2.2.1 Effect of stagger angle on cascade lift and drag characteristics ............................................... 152
CHAPTER 6 - DEVELOPMENT OF A NON-DIMENSIONAL FORM OF THE BEM

ANALYSIS . 157
6.1 NON-DIMENSIONAL MULTIPLE STREAMTUBE BEM ANALYSIS ...................................................157

6.1.1 Non-dimensional BEM model for analysis of OWC turbines ..............................................157
6.1.1.1 Tangential induction factor ...................................................................................................... 158
6.1.1.2 Pressure coefficient .................................................................................................................. 159
6.1.1.3 Input coefficient ....................................................................................................................... 159
6.1.1.4 Torque coefficient .................................................................................................................... 160
6.1.1.5 Efficiency coefficient ............................................................................................................... 161
6.1.1.6 Advantages of applying a non-dimensional BEM analysis ...................................................... 161
6.1.2 Non-dimensional multiple streamtube model ......................................................................162
6.2 ANALYSIS OF THE 1/3RD SCALE DENNISS-AULD TURBINE ..........................................................165
CHAPTER 7 - CFD ANALYSIS OF AXIAL FLOW TURBINES................................................170
7.1 HIGH SOLIDITY WELLS TURBINE MODEL WITH VARIABLE PITCH BLADES ..................................170
7.1.1 Comparison of 3D CFD and blade element analyses..........................................................177
7.2 SMALL SCALE DENNISS-AULD TURBINE MODEL........................................................................180
7.2.1 Grid independence testing ...................................................................................................180
7.2.2 Comparison of 3D CFD and experimental results ..............................................................181
viii
Table of Contents
7.2.3 Modelling of the full turbine coupled to the nacelle ............................................................183

7.3 FULL SCALE DENNISS-AULD TURBINE MODEL ..........................................................................187
7.3.2 Effect of design parameters on turbine performance...........................................................187
7.3.2.1 Tip clearance, tc ........................................................................................................................ 188
7.3.2.2 Hub to tip ratio, h ..................................................................................................................... 192
7.3.2.3 Rotor solidity, σ ....................................................................................................................... 195
7.3.2.4 Blade profile............................................................................................................................. 199
7.3.2.5 Blade thickness ratio, τ ............................................................................................................ 205
7.3.3 Adjustment of turbine design parameters ............................................................................208
7.3.4 Turbine cycle efficiency .......................................................................................................211
CHAPTER 8 - CFD ANALYSIS OF OWC PNEUMATIC SYSTEM ..........................................214
8.1 PERFORMANCE ANALYSIS OF A COMPLETE OWC PNEUMATIC SYSTEM .....................................214

8.2 ANALYSIS OF THE INTERNAL FLOW FIELD IN A NOZZLE-DIFFUSER SYSTEM ...............................219
8.2.1 Flow in horizontal air duct ..................................................................................................219
8.2.2 Flow in nozzle-diffuser in absence of the nacelle ................................................................222
8.2.3 The length of nozzle-diffuser section ...................................................................................223
8.2.4 Nacelle shape.......................................................................................................................224
8.2.5 Difference between inhaling and exhaling ..........................................................................225
CHAPTER 9 - COMPARISON OF NUMERICAL ANALYSES AND FIELD DATA FROM A

FULL SCALE OWC PLANT...................................................................................................................229
9.1 ANALYSIS OF FIELD DATA FROM A FULL SCALE OWC PLANT ....................................................229
9.2 BEM ANALYSIS OF THE FULL SCALE DENNISS -AULD TURBINE ................................................239
9.2.1 Numerical data ....................................................................................................................240
9.3 CFD ANALYSIS OF THE FULL SCALE DENNISS -AULD TURBINE .................................................243
9.3.2 Comparison with the field data............................................................................................243
9.4 COMPARISON OF THE FULL SCALE DENNISS-AULD TURBINE WITH OTHER OWC TURBINES ......246
CHAPTER 10 - CONCLUSIONS AND RECOMMENDATIONS .................................................249
10.1 CONCLUSIONS ...........................................................................................................................249

10.2 RECOMMENDATIONS FOR FURTHER RESEARCH .........................................................................251
REFERENCES ..........................................................................................................................................252
APPENDIX A-1 .........................................................................................................................................260
APPENDIX A-2 .........................................................................................................................................263
ix
Table of Contents
APPENDIX A-3 .........................................................................................................................................264
APPENDIX A-4 .........................................................................................................................................265
APPENDIX A-5 .........................................................................................................................................266
APPENDIX A-6 .........................................................................................................................................267
APPENDIX A-7 .........................................................................................................................................268
APPENDIX A-8 .........................................................................................................................................272
x
Table of Figures
TABLE OF FIGURES
Figure 2.1 Schematic showing device orientation to the wave direction (Cruz, 2008). ....8
Figure 2.2 A sketch of the cross-section of a generic shore-mounted OWC system..........9
Figure 2.3 Inclined OWC chamber of LIMPET (LIMPET, 2002). ................................11
Figure 2.4 Top view of Oceanlinx OWC device (Finnigan, 2004)..................................12
Figure 2.5 Schematic of the Wells turbine. ....................................................................13
Figure 2.6 Aerodynamic forces acting on a blade of a Wells turbine rotor.....................14
Figure 2.7 Schematic of an impulse turbine rotor with fixed GVs. .................................15
Figure 2.8 A full-scale Denniss-Auld turbine with variable pitch angle blades prior to
installation (Alcorn and Finnigan, 2004)........................................................................15
Figure 2.9 The turbine pitching regime of the Denniss-Auld turbine (Alcorn and
Finnigan, 2004). ............................................................................................................16
Figure 2.10 Power train schematic (Alcorn and Finnigan, 2004)....................................16
Figure 2.11 Scheme of the OWC Pico plant and its components (Neumann et al., 2007).
......................................................................................................................................18
Figure 2.12 OSPREY during float out from the John Brown shipyard, Glasgow, Scotland
(WaveNet, 2003). ..........................................................................................................20
Figure 2.13 Side view of sea-bottom mounted Oceanlinx OWC device (Finnigan, 2004).
......................................................................................................................................21
Figure 2.14 The world's first large-scale offshore floating prototype KAIMEI
(JAMSTEC, 2009). .......................................................................................................22
Figure 2.15 The OWC plant "Mighty Whale" (JAMSTEC, 2009)..................................22
Figure 2.16 Oceanlinx floating OWC 1/3rd scale prototype Mk2 (Oceanlinx, 2010).......23
Figure 2.17 OEBuoy (OceanEnergy, 2011) ...................................................................24
Figure 2.18 Domains and their problems, extracted from (Weber and Thomas, 2001)....26
Figure 2.19 Influence of pneumatic damping and wave height on power output of the
OWC (Curran et al., 1997b). .........................................................................................29
Figure 2.20 Influence of damping and frequency ratio on the capture factor, CF, of OWC
(Curran et al., 1997b). ...................................................................................................30
Figure 2.21 Solidity effect on damping ratio of Wells turbine (Curran et al., 1998a).....32
Figure 2.22 Performance characteristics of the monoplane and the biplane (Curran and
Gato, 1997a)..................................................................................................................34
xi
Table of Figures
Figure 2.23 Visualisation of flow separation by means of CFD (from the present study).
......................................................................................................................................35
Figure 2.24 Design parameters affecting performance of the Wells turbine...................36
Figure 2.25 The combined effect of h and AR on efficiency for Wells turbines at σ = 0.4
(Raghunathan, 1995). ....................................................................................................38
Figure 2.26 The effect of free stream turbulence on aerodynamics efficiencies
(Raghunathan, 1995). ....................................................................................................39
Figure 2.27 Variation of tangential force coefficient with the incidence (Raghunathan et
al., 1985). ......................................................................................................................40
Figure 2.28 Variation of axial force coefficient with the incidence (Raghunathan et al.,
1985).............................................................................................................................41
Figure 2.29 The effects of h and σ on the self-starting ability of the Wells turbine. The
experimental data (closed symbols) indicate conditions for self-starting (Raghunathan,
1995).............................................................................................................................42
Figure 2.30 Schematic of the Wells turbine biplane. ......................................................44
Figure 2.31 Schematic of the Wells turbine rotor with guide vanes. ..............................45
Figure 2.32 Efficiency and non-dimensional pressure drop for a monoplane configuration
w/o and with guide vanes (replotted from Curran and Gato, 1997a). ..............................46
Figure 2.33 Efficiency for a monoplane w/o guide vanes (MP), a monoplane with guide
vanes (GV), a biplane (BP) and a contra-rotating configuration (CR). ...........................47
Figure 2.34 Fixed non-zero blade angle setting, γ. Note the angle γ is measured from the
plane of blade rotation. ..................................................................................................48
Figure 2.35 Efficiency variations for a turbine without guide vanes with blades staggered
at different angles, γ (Setoguchi et al., 2002). ...............................................................49
Figure 2.36 Schematic of the biplane Wells turbine with staggered blades.....................49
Figure 2.37 Schematic of the Wells turbine with self-pitch controlled blades.................51
Figure 2.38 A schematic diagram showing the concept of the VPB turbine...................52
Figure 2.39 Partially sectioned rotor of the Pico Island Wells turbine with variable pitch
blades (Taylor and Caldwell, 1998). ..............................................................................52
Figure 2.40 Effect of blade pitch angle on the turbine efficiency (Tease, 2003)..............53
Figure 2.41 Isometric view of the symmetrical about the mid-chord blade used in the
1/3rd scale Denniss-Auld turbine tested at the University of Sydney (Finnigan and Auld,
2003), where c is the chord length and b is the blade span. ............................................54
xii
Table of Figures
Figure 2.42 Blade pitching sequence in oscillating flow (Finnigan and Auld, 2003).......54
Figure 2.43 Operating efficiency curve based on data obtained from the tests of 1/3rd
scale Denniss-Auld prototype (Finnigan and Auld, 2003). .............................................55
Figure 2.44 Effect of tip clearance, tc, on max efficiency of various turbines (Thakker and
Dhanasekaran, 2003). ....................................................................................................57
Figure 2.45 A comparison of the mean efficiencies simulated under irregular flow
conditions between the Wells turbine and the impulse turbine with guide vanes, θ =30º
(Maeda et al., 1999). .....................................................................................................58
Figure 2.46 Schematic of an impulse turbine rotor with self-pitch controlled guide vanes
(SPCGVs). ....................................................................................................................58
Figure 2.47 Conversion efficiencies of different turbine types (Kim et al., 2001)..........59
Figure 2.48 Impulse turbine configuration: a) radial and b) axial (Castro et al., 2007). ..60
Figure 2.49 HydroAir (Dresser-Rand, 2011)..................................................................61
Figure 2.50 Rotating element of fluid (Wallis, 1961). ....................................................63
Figure 2.51 Annular plane used in BEM theory. ............................................................65
Figure 2.52 An example of the computational domain with rotational periodic
boundaries.....................................................................................................................66
Figure 2.53 Computational domain for VARC rotor with guide vanes (Govardhan and
Chauhan, 2007). ............................................................................................................69
Figure 2.54 Variation of power coefficient, W*, with φ for CONC and VARC rotors
(Govardhan and Chauhan, 2007). ..................................................................................70
Figure 2.55 Effect of guide vanes on efficiency, η, of VARC rotor (Govardhan and
Chauhan, 2007). ............................................................................................................70
Figure 2.56 Two stage rotor setup with two rotating blade sections and one stationary
generator section with fixation of generator (Arlitt et al., 2007).....................................71
Figure 3.1 Actuator disk and mass flow rate balance for a ducted turbine. .....................74
Figure 3.2 Blade element velocity vectors. ...................................................................76
Figure 3.3 Forces acting on the blade............................................................................76
Figure 3.4 ANSYS CFX software modules that pass the information required to perform
a CFD analysis (ANSYS CFX Introduction, 2006). .......................................................79
Figure 3.5 Representation of near-wall region (Cengel and Cimbala, 2006). ..................87
Figure 3.6 Example of computational domain with rotational periodicity for a Wells
turbine model (γ = 24°)..................................................................................................88
xiii
Table of Figures
Figure 3.7 Structured elements generated with aid of TuboGrid by the present author for
the Wells turbine with blades staggered at 32° relative to the plain of rotation. ..............90
Figure 3.8 An example of the complex geometry of the pneumatic system of an OWC
WEC including the turbine coupled to the nacelle meshed by the unstructured elements
(made by the present author)..........................................................................................90
Figure 3.9 Example of a combination unstructured and structured grid elements around a
cascade of NACA0012 aerofoils....................................................................................91
Figure 3.10 The example of the mesh control of the region with anticipated high
gradients........................................................................................................................93
Figure 3.11 A flowchart presenting the flow physics according to the research objectives.
......................................................................................................................................94
Figure 3.12 Setting up the rotational periodic and general connection interfaces for 3D
CFD modelling of the Wells turbine with blades staggered at 24°..................................95
Figure 3.13 An overview of the solution procedure........................................................97
Figure 3.14 A typical 3D view of the Wells turbine model with a blade-to-blade contour
plot showing the pressure distribution through the computational domain at the blade
mid-span. ......................................................................................................................98
Figure 4.1 Lift coefficients over the range of α from 0º to 32º for Re of around 2–3
million for NACA0012 airfoil (Gretton and Bruce, 2007)............................................ 101
Figure 4.2 Measured CL curve compared to XFOIL free transition calculation and
EllipSys2D turbulent flow calculation at Re = 1.3×106 (Fuglsang et al., 1998). ............ 102
Figure 4.3 Measured CD curve compared to XFOIL free transition calculation and
EllipSys2D turbulent flow calculation at Re = 1.3×106 (Fuglsang et al., 1998). ............ 103
Figure 4.4 A comparison of CL for NACA 0012:(Jacobs and Sherman, 1937) at
Re=6.6×105; (Sheldahl and Klimas, 1981) at Re=7×105; (Loftin and Smith, 1949) at
Re=7×105. ................................................................................................................... 104
Figure 4.5 A comparison of CL for NACA 0012:(Loftin and Smith, 1949) at Re=2×106;
(Sheldahl and Klimas, 1981) at Re=2×106; (Critzos et al., 1955) at Re=1.8×106. ......... 104
Figure 4.6 A comparison of CL coefficients for NACA 0012:(Critzos et al., 1955) at
Re=1.8×106; (Sheldahl and Klimas, 1981) at Re=2×106. .............................................. 105
Figure 4.7 A comparison of CD coefficients for NACA 0012:(Critzos et al., 1955) at
Re=1.8×106; (Sheldahl and Klimas, 1981) at Re=2×106. .............................................. 105
Figure 4.8 Dependence of CL coefficient from the number of elements, n. ................... 107
xiv
Table of Figures
Figure 4.9 Dependence of CD coefficient from the number of elements, n.................... 108
Figure 4.10 A comparison between experimental CL data for an isolated NACA64A010
aerofoil (open symbols/solid line) published by Selig et al., (1995) and CFD results
(symbols: □-k-ε model, ●-SST model) obtained by the present author. ......................... 109
Figure 4.11 A comparison between experimental CD data for an isolated NACA64A010
(symbols: □-k-ε model, ●-SST model) obtained by the present author. ......................... 110
Figure 4.12 A comparison between experimental CL data for an isolated NACA6409
(closed symbols) obtained by the present author. ......................................................... 111
Figure 4.13 A comparison between experimental CD data for an isolated NACA6409
aerofoil (solid line with open symbols) published by Selig et al., (1995) and CFD results
(closed symbols) obtained by the present author. ......................................................... 112
Figure 4.14 A comparison of CL data for NACA0012 aerofoils at Re = 3.3×105 (Jacobs
and Sherman, 1937) and Re = 3.6×105 (Sheldahl and Klimas, 1981) and CFD analysis at
Re = 3.6×105 in the present study. ............................................................................... 114
Figure 4.15 A comparison of CD data for NACA0012 aerofoils at Re = 3.6×105 (Sheldahl
and Klimas, 1981) and CFD analysis at Re = 3.6×105 in the present study. .................. 115
Figure 4.16 A 1/3rd scale prototype of the Denniss-Auld turbine design (Finnigan and
Auld, 2003). ................................................................................................................ 116
Figure 4.17 CL and CD coefficients, as measured in wind tunnel tests (-5°< α <18°), and
as modelled (Finnigan and Alcorn, 2003). ................................................................... 117
Figure 4.18 The CL coefficients for an aerofoil used in the 1/3rd scale Denniss-Auld
turbine obtained in present study (closed symbols/ broken line) and lift data (open
symbols/solid line) reported by Finnigan and Alcorn (2003)........................................ 119
Figure 4.19 The CD coefficients for an aerofoil used in the 1/3rd scale Denniss-Auld
turbine obtained in present study (closed symbols/broken line) and lift data (open
symbols/solid line) reported by Finnigan and Alcorn (2003)........................................ 120
Figure 4.20 Illustration of recirculation behind the aerofoil section positioned at α =
90°.The aerofoil profile used in the 1/3rd scale Denniss-Auld turbine........................... 121
Figure 4.21 A circular arc profile blade as used in the full scale Denniss-Auld turbine.122
Figure 4.22 The dimensions of a domain for modelling of flow over an isolated circular
arc profile aerofoil at α = -80°. .................................................................................... 122
xv
Table of Figures
Figure 4.23 The CL (open symbols/solid line) and CD (closed symbols/broken line) data
for a circular arc profile aerofoil obtained by CFD in present study at Re = 2.5×105. ... 124
Figure 4.24 Variation of CL coefficients with angle of attack for a cambered plate
(Pandey et al., 1988).................................................................................................... 125
Figure 4.25 A comparison between experimental (Pandey et al., 1988) and CFD (present
study) lift data for 0° ≤ α ≤ 90°. .................................................................................. 125
Figure 4.26 Illustration of recirculation behind the circular arc profile aerofoil positioned
at α = -80°................................................................................................................... 126
Figure 5.1 Nomenclature for a cascade of aerofoils with stagger angle, γ, on an axial flow
turbine rotor, with axial air velocity, VA, and tangential velocity relative to aerofoils, VT.
.................................................................................................................................... 129
Figure 5.3 Prediction of the interference factor, k0 = CL/CL0, from potential flow theory
(Weinig, 1964). γeff = 90° - γ, where the γeff is the stagger angle of the flat plates
measured from the axis of X........................................................................................ 131
Figure 5.4 Representation of isolated lift data CL0 (solid line) based on Sheldahl and
Klimas (1981) at Re = 7×105 and predicted cascade lift data CL corrected by k0 from (5.2)
and assuming s/c = 1.69............................................................................................... 132
Figure 5.5 Nomenclature for a cascade of aerofoils with stagger angle, γ = 0°, on an axial
flow turbine rotor with chord length, c, and pitch between blades, s; αm = (α1 + α2)/2;
Wm= (W1+W2)/2........................................................................................................... 134
Figure 5.6 Normalised values of CL for a NACA0012 aerofoil at different solidities
(Wong, 1994). ............................................................................................................. 135
Figure 5.7 Normalised values of Cx for a NACA0012 aerofoil at different solidities
(Wong, 1994). ............................................................................................................. 135
Figure 5.8 CFD prediction of streamlines in a linear and infinite cascade of NACA0012
aerofoils (γ = 0º) with s/c=2 and αm = 20º. ................................................................... 137
Figure 5.9 Lift coefficient interference factor, k0, for a linear cascade (γ = 0º) of
NACA0012 aerofoils as a function of s/c and the mean angle of incidence, αm: □ 5º; ∆
10º; ◊ 15º; ○ 20º. Broken lines represent CFD prediction; solid line with * symbols is
based on Weinig’s inviscid flow analysis, Eq. (5.2). .................................................... 138
Figure 5.10 Pressure distribution for: a) s/c=1.25 and b) s/c=2 at αm = 20º................... 139
Figure 5.11 Interference factor for drag, δ0, deduced from CFD simulations
complementing the results shown in Figure 5.8. .......................................................... 139
xvi
Table of Figures
Figure 5.12 Lift coefficient interference factor (k0) for a linear cascade (γ = 0º) of
based on Weinig’s inviscid flow analysis, Eq. (5.2). .................................................... 140
Figure 5.13 CFD prediction of cascade lift data (solid line with opened symbols),
complementing the results shown in Figure 5.3 for s/c = 1.69. .................................... 141
Figure 5.14 Normalised CL for a linear cascade (γ = 0º) of NACA0012 aerofoils as a
function of solidity, σ =c/s, and the mean angle of incidence, αm: □ 5º; ∆ 10º; ◊ 15º; ○
20º. Broken lines represent CFD prediction; solid line with *symbols is based on Wong’s
inviscid flow analysis (Martensen method). ................................................................. 142
Figure 5.15 Normalised Cx for a linear cascade (γ = 0º) of NACA0012 aerofoils as a
function of solidity, σ = c/s, and the mean angle of incidence, αm: □ 5º, ∆ 10º, ◊ 15º, ○
20º. Broken lines represent CFD prediction; solid line with * symbols is based on Wong’s
inviscid flow analysis (Martensen method); solid line with × symbols represents the
experimental data for αm= 5° (Raghunathan et al., 1985). ............................................ 143
Figure 5.16 Comparison of experimental data of CL for cascades of three and five
NACA0021 aerofoils (Raghunathan, 1988) with a CFD simulation from the present work
for an infinite cascade (c/s = 0.5). ................................................................................ 144
Figure 5.17 Comparison of experimental data of CD for cascades of three and five
for an infinite cascade (c/s = 0.5). ................................................................................ 145
Figure 5.18 Comparison of lift coefficients deduced from CFD analysis and experimental
rotor tests (Curran et al., 1998a) for a monoplane Well turbine with NACA0012 blade
profile (σ = 0.64). Isolated aerofoil lift data are also provided (Sheldahl and Klimas,
1981)........................................................................................................................... 146
Figure 5.19 Comparison of drag coefficients (for same conditions as in Figure 5.17)... 147
Figure 5.20 CFD prediction of lift interference factor for linear cascade of blades similar
to those of (Finnigan and Auld, 2003) for an upstream angle of incidence α = 10º. Note in
this figure γ′ is measured from the plane normal to the plane of rotation. ..................... 148
Figure 5.21 Interference factor for drag, δ0, complementing the results shown in Figure
5.19. Note in this figure γ′ is measured from the plane normal to the plane of rotation. 148
Figure 5.22 Comparison between Weinig’s data (broken lines) and CFD results (solid
lines) for aerofoils staggered at 0° and 70°................................................................... 149
xvii
Table of Figures
Figure 5.23 Cascade interference factor, k1 for circular arc profile with smooth inflow
conditions (Weinig, 1964). Note in this figure γi is measured from the plane normal to the
plane of rotation. ......................................................................................................... 150
Figure 5.24 Comparison between Weinig’s data shown in Figure 5.22 and CFD results
for two linear cascades of the full scale Denniss-Auld aerofoils for an upstream angle of
incidence α = 0º. CFD results for s/c = 0.9: □ 20º, ∆ 30º, ◊ 35º, ○ 40º, * 45º; CFD results
for s/c = 1.5: □ 20º, ∆ 30º, ◊ 35º, ○ 40º, * 45º, +50º, × 60º, ▲ 75º, ■ 80º. Note in this
figure γ′ is measured from the plane normal to the plane of rotation............................. 152
Figure 5.25 Effect of stagger angle on cascade lift data (c/s = 0.49). Isolated airfoil lift
data tested at comparable Reynolds number are also shown for comparison (Sheldahl and
Klimas, 1981). Note in this figureγ is measured from the plane of rotation. ................. 153
Figure 5.26 Effect of stagger angle on cascade lift data (c/s = 0.49) Isolated airfoil lift
Klimas, 1981). Note in this figureγ is measured from the plane of rotation. ................. 154
Figure 6.1 Four streamtubes air duct representation. .................................................... 162
Figure 6.2 Four streamtubes blades representation....................................................... 163
Figure 6.3 Numerical results (solid lines) from the present 10 streamtube BEM analysis
(broken lines) for the efficiency of a laboratory-scale Denniss-Auld turbine (isolated
aerofoil lift/drag data deduced from CFD simulations) compared with experiments (□
γ =20º, ○ γ = 40º, ∆ γ = 60º, ◊ γ =80º) by Finnigan and Auld (2003). Results based on
single streamtube BEM model (solid lines) are also given for comparison. .................. 166
Figure 6.4 Torque coefficient, CT, predicted from isolated lift/drag data deduced from
CFD simulations. Note in this figure γ is measured from the plane of rotation. ............ 167
Figure 6.5 Input coefficient, CA, predicted from isolated lift/drag data deduced from CFD
simulations. ................................................................................................................. 168
Figure 6.6 Pressure coefficient, P*, predicted from isolated lift/drag data deduced from
CFD simulations.......................................................................................................... 168
Figure 7.1 Schematic of the turbine test rig (Tease, 2003)............................................ 172
Figure 7.2 Computational domain used for modelling of the Wells turbine reported by
Tease (2003) with thirteen NACA0012 blades staggered at γ = 3º and σ =0.48............ 172
Figure 7.3 Streamlines over a Wells turbine blade with γ = 3° at different flow
coefficients, φ. ............................................................................................................. 173
xviii
Table of Figures
Figure 7.4 Distribution of streamlines in vicinity of a Wells turbine blade with γ = 3° at

different flow coefficients, φ (complementing Figure 7.3). .......................................... 173
coefficients, φ. ............................................................................................................. 174
Figure 7.6 Distribution of surface streamlines on the suction side of the Wells turbine
blade with γ = 24° at different flow coefficients, φ, (complementing Figure 7.5).......... 175
Figure 7.7 Results from the present 3D CFD simulations (closed symbols/broken lines)
for non-dimensional pressure, P*, as a function of flow factor, φ, for different stagger
angles compared with the experimental results (open symbols) reported by Tease (2003).
.................................................................................................................................... 175
coefficients, φ. ............................................................................................................. 176
blade with γ = 32° at different flow coefficients, φ, (complementing Figure 7.8).......... 176
Figure 7.10 Variation of torque coefficient, CT, from 3D CFD simulations by the present
author of a variable pitch angle Wells turbine (Tease, 2003). ....................................... 177
Figure 7.11 Comparison of pressure coefficient results from 3D CFD simulations (closed
symbols/broken lines) and from the BEM 10-streamtube model (open symbols/solid
lines), for the geometry of the Wells turbine reported by Tease (2003). ....................... 178
Figure 7.12 Comparison of numerical results based on a blade element model (open
symbols/solid lines) and isolated lift/drag data taken from Sheldahl and Klimas (1981)
against results for CT deduced from 3D CFD simulations (closed symbols/broken lines).
Both set of data obtained in present study for a Wells turbine reported by Tease (2003).
.................................................................................................................................... 179
Figure 7.13 3D CFD results (closed symbols/solid lines) from the present study for the
efficiency of a 1/3rd scale Denniss-Auld turbine model compared with experiments (open
symbols: □ γ =20º, ○ γ = 40º, ∆ γ = 60º, ◊ γ =80º) by Finnigan and Auld (2003). Note in
this figureγ is measured from the plane of rotation....................................................... 181
Figure 7.14 Predicted torque coefficient, CT, (opened symbols/solid lines) compared with
experiments (open symbols: □ γ =20º, ○ γ = 40º, ∆ γ = 60º, ◊ γ =80º) by Finnigan and
Auld (2003). Note in this figure γ is measured from the plane of rotation..................... 182
xix
Table of Figures
Figure 7.15 Predicted input coefficient, CA, closed symbols/solid lines) compared with
Auld (2003). Note in this figureγ is measured from the plane of rotation...................... 183
Figure 7.16 Isometric view of a 1/3rd scale Denniss-Auld turbine model with blades
staggered at 20º. The blade rotation is shown by the red arrow. ................................... 185
Figure 7.17 Comparison between efficiencies, η, deduced from 3D CFD modelling of a
full 1/3rd scale Denniss-Auld turbine (open symbols/solid lines) and a 1/8th of the turbine
(closed symbols/broken lines). Experimental data (symbols: ▲20°,♦80°)................... 185
Figure 7.18 Comparison between torque coefficients, CT, deduced from 3D CFD
modelling of a full 1/3rd scale Denniss-Auld turbine (open symbols/solid lines) and a 1/8th
of the turbine (closed symbols/broken lines). Experimental data (symbols: ▲20°,♦
♦80°).
.................................................................................................................................... 186
Figure 7.19 Comparison between input coefficients, CA, deduced from 3D CFD
of the turbine (closed symbols/broken lines). Experimental data (symbols: ▲20°, ◊ 80°).
.................................................................................................................................... 186
Figure 7.20 Typical meshing of the tip gap (TG) generated by TurboGrid. .................. 188
Figure 7.21 Effect of tip clearance, tc, on efficiency, η, of the full scale Denniss-Auld
turbine model with blades staggered at 45°, 55° and 70°.............................................. 189
Figure 7.22 Effect of tip clearance, tc, on torque coefficient, CT, of the full scale Denniss-
Auld turbine model with blades staggered at 45°, 55° and 70°. .................................... 190
Figure 7.23 Effect of tip clearance, tc, on pressure coefficient, P*, of the full scale
Denniss-Auld turbine model with blades staggered at 45°, 55° and 70°. ...................... 190
Figure 7.24 Surface streamlines over the suction side of blades with tc =1% and tc = 3%
staggered at 45°. The flow coefficient, φ, was equal to 2.16......................................... 191
Figure 7.25 Effect of hub to tip ratio, h, on efficiency of the full scale Denniss-Auld
turbine model with blades staggered at 45º and 70º...................................................... 193
Figure 7.26 Effect of hub to tip ratio, h, on torque characteristic of the full scale Denniss-
Auld turbine model with blades staggered at 45º and 70º. ............................................ 194
Figure 7.27 Effect of hub to tip ratio, h, on pressure characteristic of the full scale
Denniss-Auld turbine model with blades staggered at 45º and 70º. .............................. 194
xx
Table of Figures
Figure 7.28 Efficiency and pressure coefficient curves for Wells turbine published by
Gato and Falcao (1988). Symbols show experimental results and lines are from their
numerical analysis. The parameter U* represents the flow coefficient, φ. ..................... 196
Figure 7.29 Effect of rotor solidity, σ, on efficiency, η, of the turbine model with blades
staggered at 45º. .......................................................................................................... 197
Figure 7.30 Effect of rotor solidity, σ, on torque characteristic, CT, of the turbine model
with blades staggered at 45º......................................................................................... 198
Figure 7.31 Effect of rotor solidity, σ, on pressure coefficient, P*, of the turbine model
with blades staggered at 45º......................................................................................... 198
Figure 7.32 Type C profile. ........................................................................................ 199
Figure 7.33 Type D profile. ........................................................................................ 199
Figure 7.34 Effect of blade profile on efficiency, η, of Denniss-Auld turbine. Type C
(closed symbols/solid lines) and type D (closed symbols/broken lines)........................ 200
Figure 7.35 Effect of blade profile on input coefficient, CA, of Denniss-Auld turbine.
Type C (closed symbols/solid lines) and type D (closed symbols/broken lines). .......... 201
Figure 7.36 Effect of blade profile on torque coefficient, CT, of Denniss-Auld turbine.
Type C (closed symbols/solid lines) and type D (closed symbols/broken lines). .......... 202
Figure 7.37 Variations of α at which the resultant velocity, WR, acted on the blades. ... 202
Figure 7.38 Isometric views of the pressure surfaces of C and D blades staggered at 45º at
different flow coefficients, φ........................................................................................ 203
Figure 7.41 Effect of blade profile on efficiency characteristics of the Wells turbine: (a)
steady-state, (b) sinusoidal flow conditions (Setoguchi et al., 2003a)........................... 205
Figure 7.42 Type C profile: blade thickness ratio, τ = 0.049. ...................................... 205
Figure 7.43 Type C profile: blade thickness ratio, τ = 0.06 (baseline). ......................... 205
Figure 7.44 Type C profile: blade thickness ratio, τ = 0.073. ....................................... 206
Figure 7.45 Effect of τ on efficiency of turbine model with C blades staggered at 45° and
70°. Baseline curves (closed symbols/solid lines). ....................................................... 206
xxi
Table of Figures
Figure 7.46 Effect of τ on CT of turbine model with C blades staggered at 45° and 70°.
Baseline curves (closed symbols/solid lines)................................................................ 207
Figure 7.47 Effect of τ on CA of turbine model with C blades staggered at 45° and 70°.
Baseline curves (closed symbols/solid lines)................................................................ 207
Figure 7.48 A comparison between efficiencies, η, based on the original Denniss-Auld
turbine geometry and a turbine with the adjusted configuration. .................................. 209
Figure 7.49 A comparison between input coefficients, CA, based on the original Denniss-
Auld turbine geometry and a turbine with the adjusted configuration........................... 210
Figure 7.50 A comparison between torque coefficients, CT, based on the original Denniss-
Auld turbine geometry and a turbine with the adjusted configuration........................... 210
Figure 8.1 Oceanlinx OWC plant tested at Port Kembla, Australia. ............................. 214
Figure 8.2 Isometric view of a full scale Denniss-Auld turbine with blades staggered at
45º incorporated in OWC wave energy converter (exhalation cycle)............................ 216
Figure 8.3 Typical pressure contour during exhalation cycle. The mass flow rate through
the rotor was 33.5kg/s and air density was 1.20kg/m3. ................................................. 217
Figure 8.4 Typical pressure gradient diagram of the OWC during exhaling (+) and
inhaling (-) cycles. The mass flow rate through the rotor was 33.5kg/s. ....................... 217
Figure 8.5 Relative total pressure loss in OWC wave energy converter during a) exhaling
and b) inhaling (mass flow rate = 33.5kg/s). ................................................................ 218
Figure 8.6 Pneumatic energy conversion/loss in a typical OWC wave energy converter
(similar to the Oceanlinx Plant). .................................................................................. 218
Figure 8.7 Arrangement of Turbo Generation Equipment (LIMPET, 2002) ................. 220
Figure 8.8 Velocity flow field in a horizontal air duct 1.64m in diameter with a nacelle
1.2m in diameter.......................................................................................................... 220
1.2m in diameter.......................................................................................................... 221
Figure 8.10 Velocity flow field in converging nozzle section with included angle of 7°.
Diameter at the outlet is 1.57m. Nacelle diameter is 1.2m............................................ 221
Figure 8.11 Velocity flow field in the horizontal nozzle-diffuser comprising the
converging and diverging sections with included angle of 9°. Velocity at the inlet was
2.5m/s. ........................................................................................................................ 222
Figure 8.12 Illustration of the flow separation from the lower side of the diverging part of
the air duct. Complimenting results shown in Figure 8.20............................................ 223
xxii
Table of Figures
Figure 8.13 Effect of nozzle-diffuser length on pneumatic efficiency, ηpneu of OWC

system. ........................................................................................................................ 224
Figure 8.14 Velocity flow field in horizontal nozzle-diffuser with the baseline nacelle
shape (no rotor is presented). Velocity at the inlet to base of the OWC chamber was
1.02m/s. ...................................................................................................................... 225
Figure 8.15 Velocity flow field in horizontal nozzle-diffuser with the nacelle having
conical noses (no rotor is presented). Velocity at the inlet to base of the OWC chamber
was 1.02m/s................................................................................................................. 225
Figure 8.16 Axial velocity field in a nozzle-diffuser during inhaling flow cycle. The
blades were modelled as rotating at 350RPM with stagger angle, γ = 45°. ................... 226
Figure 8.17 Axial velocity field in nozzle-diffuser during exhaling flow cycle. The blades
were modelled as rotating at 350RPM with stagger angle, γ = 45°. .............................. 227
Figure 9.1 Sensor layout (provided by Oceanlinx). ..................................................... 230
Figure 9.2 An example of comparison of volume flowrates, Q, calculated using data from
various pressure sensors and hot-film anemometer in the full scale Oceanlinx OWC Port
Kembla plant. The numbers from 2 to 9 represent the pressure sensors........................ 232
Figure 9.3 An example of comparison of different turbine characteristics as a function of
time (the power threshold is 10kW, i.e. efficiency only calculated for instances where this
power threshold was exceeded). .................................................................................. 233
Figure 9.4 Specific examples of instantaneous efficiency, η, as a function of flow factor,
φ, and the blade stagger angle of +/-45° during a number of wave peaks/troughs (length
of data record is 660s, power threshold is 10kW). Closed symbols are periods of
exhalation and open symbols are during inhalation. ..................................................... 235
Figure 9.5 Instantaneous pressure coefficient, P*, as a function of flow factor, φ, and the
blade stagger angle of +/-45° (length of data record is 660s, power threshold is 10kW).
Closed symbols are periods of exhalation and open symbols are during inhalation. ..... 235
Figure 9.6 Instantaneous torque coefficient, CT, as a function of flow factor, φ, and the
Closed symbols are periods of exhalation and open symbols are during inhalation. ..... 236
Figure 9.7 Instantaneous efficiency, η, as a function of flow factor, φ, and the blade
stagger angle, γ, (length of data record is 660s, power threshold is 10kW). ................. 237
xxiii
Table of Figures
blade stagger angle, γ, (length of data record is 660s, power threshold is 10kW). ........ 238
blade stagger angle, γ, (length of data record is 660s, power threshold is 10kW)......... 238
Figure 9.10 Geometrical parameters and technical features of the Denniss-Auld turbine
(made by present author). ............................................................................................ 240
Figure 9.11 Instantaneous efficiency (symbols), η, as a function of φ and γ compared to
the results based on the BEM analysis (solid lines). ..................................................... 241
Figure 9.12 Instantaneous pressure coefficient (symbols), P*, as a function of φ and γ
compared to the results based on the BEM analysis (solid lines). ................................. 242
Figure 9.13 Instantaneous torque coefficient (symbols), CT, as a function of φ and γ
compared to the results based on the BEM analysis (solid lines). ................................. 242
Figure 9.14 Instantaneous efficiency (symbols/broken lines), η, as a function of flow
factor, φ, from full scale tests of the Denniss-Auld turbine at Port Kembla compared to
the results of 3D CFD simulations (open symbols/solid lines) obtained for γ = 45º, 55º
and 70º. Curves from the BEM analysis are given for comparison (broken lines)......... 244
Figure 9.15. Instantaneous efficiency (symbols/broken lines), η, as a function of flow
the results of 3D CFD simulations (open symbols/solid lines) obtained for γ = 50º and
60º. Curves from the BEM analysis are given for comparison (broken lines). .............. 245
Figure 9.16 Instantaneous pressure coefficient (symbols), P*, as a function of flow factor,
φ, compared to the results of 3D CFD simulations (closed symbols/solid lines). Curves
from the BEM analysis are given for comparison (broken lines). ................................. 245
Figure 9.17 Instantaneous torque coefficient (symbols), CT, as a function of flow factor,
φ, compared to 3D CFD simulations (closed symbols/solid lines). Curves from the BEM
analysis are given for comparison (broken lines). ........................................................ 246
xxiv
List of Tables
LIST OF TABLES
Table 2.1 Major OWC turbines tested in real sea conditions ..........................................61
Table 4.1 Mesh refinement settings for an aerofoil used in the 1/3rd scale Denniss-Auld
turbine......................................................................................................................... 117
Table 4.2 Grid independence tests of a aerofoil used in the 1/3rd scale Denniss-Auld
turbine......................................................................................................................... 118
Table 4.3 Mesh refinement settings for a circular arc profile aerofoil used in the full scale
Denniss-Auld turbine .................................................................................................. 123
Table 4.4 Grid independence testing of a circular arc profile blade used in the full scale
Denniss-Auld turbine .................................................................................................. 123
Table 5.1 The calculated results of k0 based on Weinig’s relationship, Eq. (5.2). ......... 137
Table 7.1 Details of the Wells turbine with variable pitch blades reported by Tease (2003)
.................................................................................................................................... 171
Table 9.1 Summary of the performance characteristics of self-rectifying turbines
discussed in the present study. ..................................................................................... 247
xxv
Nomenclature
NOMENCLATURE
A = Planform area of the aerofoil section [m2]

AA = Turbine annular duct area [m2]
AB = Total blade area [m2]
AC = Free surface area of OWC chamber [m2]
Ad = Cross sectional area of the duct [m2]
AD = Actuator disk area [m2]
AR = Column-duct area ratio [--]
Aw = Downstream streamtube area [m2]
A∞ = Upstream streamtube area [m2]
AR = Aspect ratio of the blade [--]
Ar = Area ratio of the diffuser [--]
a = Axial flow induction factor [--]
a′ = Tangential flow induction factor [--]
B = Damping [Ns/m]
BA = Applied damping [Ns/m]
B2 = Secondary damping [Ns/m]
BR = Damping ratio [--]
b = Blade span [m]
C = Log-layer constant [--]
CA = Input coefficient [--]
CD = Drag coefficient [--]
Cd = Discharge coefficient [--]
CL = Lift coefficient [--]
CL0 = Lift coefficient for an isolated aerofoil [--]
CLcs = Lift coefficient for cascade flow [--]
CT = Torque coefficient [--]
Cθ = Tangential blade force coefficient [--]
Cθ0 = Tangential force coefficient for isolated aerofoil [--]
Cx = Axial blade force coefficient [--]
Cx0 = Axial force coefficient for isolated aerofoil [--]
Cµ = k-ε turbulence model constant [--]
xxvi
Nomenclature
c = Chord length of blade [m]

ca = Speed of sound [m/s]
D = Pipe diameter [m]
DH = Hub diameter [m]
DT = Blade tip diameter of the rotor [m]
d = Diameter of the orifice [m]
dr = Thickness of a streamtube [m]
F = Force [N]
FA = Axial force on aerofoil (blade) [N]
Fc = Centrifugal force [N]
FD = Drag force [N]
FL = Lift force [N]
Fp = Pressure force [N]
FT = Tangential force on aerofoil (blade) [N]
f = Concavity of aerofoil [--]
G = Gap between rotors [m]
g = Coefficient of gravitational acceleration [m/s2]
H = Total pressure of fluid element [Pa]
h = Hub-to-tip ratio [--]
I = Moment of Inertia of Rotor [kg.m2]
K = Restoring constant of the spring [--]
k = Dimensionless period [--]
k0 = Cascade interference factor for Lift [--]
k1 = Cascade interference factor for circular arc [--]
L = Length [m]
M = Mach number [--]
Mcrt = Critical Mach number [--]
ME = Total Mass of the OWC [kg]
m = Mass of body [kg]
m
& = Mass flow rate of air [kg/s]
N = Number of blades [--]
Np = Number of planes [--]
n = Number of streamtubes [--]
xxvii
Nomenclature
P* = Non-dimensional pressure drop across turbine [--]

Pa = Atmospheric pressure [Pa]
PD = Pressure across actuator disc [Pa]
Pk = Production rate of turbulence [--]
Pst = Static pressure [Pa]
P∞ = Free stream pressure [Pa]
P = Oscillating air pressure [Pa]
∆P = Pressure drop across the turbine [Pa]
∆P0 = Total pressure drop [Pa]
∆P0 = Total pressure drop [Pa]
∆y = Distance from the wall [m]
p = Pressure [Pa]
pref = Reference pressure [Pa]
Q = Axial flow rate [m3/s]
q = Volume flow rate of air [m3/s]
qd = Diffraction air flow rate [m3/s]
qr = Radiation flow rate [m3/s]
R = Gas constant [J/K mol]
R0 = Universal gas constant [J/K mol]
RAV = Average radius of the blades [m]
Re = Reynolds Number [--]
RH = Hub radius [m]
RT = Tip radius of the rotor [m]
r = Radius [m]
s = Pitch (spacing) between blades [m]
s* = Shift between planes [m]
s′ = Size of element [m]
T = Thermodynamic temperature [K°]
To = Total torque of the rotor [Nm]
Tu = Turbulence level [--]
t = Time [s]
tc = Tip clearance [m]
to = Torque per unit length [Nm]
xxviii
Nomenclature
U = Flow velocity [m/s]

u+ = Near wall velocity [m/s]
uτ = Friction velocity [m/s]
UD = Actuator disk velocity [m/s]
Uw = Downstream velocity [m/s]
U∞ = Free stream velocity [m/s]
V = Volume inside the OWC chamber [m3]
V0 = Undisturbed value of V [m3]
V = Fluid velocity [m/s]
VA = Axial velocity across the turbine [m/s]
V1 = Velocity through the pipe with diameter of D [m/s]
V2 = Velocity at the orifice with diameter of d [m/s]
VC = Air velocity in the OWC chamber [m/s]
Vcart = Velocity based on Cartesian coordinates [m/s]
Vnor = Normal speed [m/s]
VT = Tangential velocity of the blade [m/s]
WR = Resultant relative velocity of the blade [m/s]
WOWC = Average pneumatic power output [W]
Wt = Turbine power output [W]
+
y = Dimensionless distance from the wall [--]
Greek Letters
α = Angle of attack [deg]
αeq = Angle of equivalent incidence [deg]
αm = Mean angle of attack [deg]
αs = Stall angle [deg]
β = Angle of incidence at WR [deg]
δ0 = Cascade interference coefficient for Drag [--]
φ = Flow factor [--]

φ = Diameter ratio [--]
γ = Stagger angle of the blade [deg]
γ′ = Stagger angle (Weinig’s definition (γ ′=90°-γ) [deg]
xxix
Nomenclature
θ = Setting angle of the GVs [deg]
θ′ = Included angle of the diffuser [deg]
η = Turbine efficiency [--]
ηD = Diffuser efficiency [--]
ηm = Conversion efficiency [--]

η = Mean efficiency [--]
µ = Dynamic viscosity coefficient [N s/m2]
µeff = Effective dynamic viscosity coefficient [N s/m2]
µt = Turbulence viscosity coefficient [N s/m2]
ν = Kinematic viscosity coefficient [m2/s]
τ = Aerofoil (blade) thickness ratio [--]

τω = Wall shear stress [Pa]
ρ = Fluid density [kg/m3]
ρa = Density of air [kg/m3]
ρw = Density of water [kg/m3]
Ω = Rotational speed of the rotor [rad/s]

σ = Turbine solidity [--]
ω = Rotational speed of turbine [rad/s]
ωOWC = Angular frequency of oscillation [rad/s]
ωO = Resonant frequency [rad/s]
ω* = Dimensionless angular velocity [--]
ω = Turbulent frequency [--]

Φ = Non-dimensional flow rate [--]
Ψ = Non-dimensional pressure [--]
Π = Non-dimensional turbine power output [--]
∇ = Vector operator [--]
xxx
Chapter 1
CHAPTER 1 - INTRODUCTION
Energy resources are vital for the wellbeing of modern civilisation. A key issue today
is the supply of energy needed by the growing economies of the world. Numerous studies
point out that there will likely be a doubling of the world’s energy demand by 2050
(WEC, 2007). It is obvious that humankind must adequately respond to this increasing
energy demand. To further rely on traditional methods of energy production based on the
burning of fossil fuels is not possible in the long term. This is not only because fossil
fuels are limited, but also because of their significant contribution to Global Warming.
The urgent need for large-scale and pollution-free power generation alternatives is a
major issue on the agenda of modern civilization. Renewable energy is one of the
alternatives being increasingly put to commercial use around the world. The range of
recognised renewable energy sources is spreading beyond well established technologies
such solar, wind to hydro-power. Ocean waves are one of the most promising hydro-
power resources for emission-free energy, which can be utilized by humankind in the
medium and long term. An important feature of ocean waves is their high energy density,
which is the highest among renewable power sources. It should be mentioned that for
many countries, security and diversity of energy supply are part of the driver for offshore
energy research and production.
Attempts to extract power from ocean waves have been undertaken by humankind for
a long time. One of the first methods for the extraction of wave energy was patented in
1799 (Clement et al., 2002). Since then time references in the technical literature to ideas
that describe the development of wave power conversion devices have been carefully
documented from the first British patent in 1855 up to 1973, when there were already 340
patents (Clement et al., 2002).
However, the greatest attention to energy production from ocean waves has been
demonstrated within the last four decades since the first oil crisis in 1973. In response to
the dramatic increase in oil prices in 1973 a dramatic increase in research and
development of wave energy conversion systems was initiated with support of
government funding in several developed countries including UK, Ireland, Norway,
Portugal, Denmark, Sweden and Japan. Among the main aims of this process were to
deliver projects focussed on medium to large scale implementation. By the mid 1980s a
significant number of wave energy conversion methods have been invented and some
novel concepts were developed to the stage of prototype construction.
1
Chapter 1
Nevertheless, the goals, which were set in the mid 1970s in regards to
commercialisation of large scale prototypes, were not achieved and governmental
funding was reduced considerably in the mid 1980s.
With the advent of a new millennium we see again the rising interest in the field of
renewable energy alternatives, which continues to increase after formulation of the Kyoto
Protocol in 1997. Growing concern regarding climate change gave a new impulse to
research and development of renewable power technologies, opening new horizons in
abundant and pollution-free ocean wave energy for humankind.
Over recent years a range of new medium to full scale projects targeting ocean wave
energy have commenced under governmental funding worldwide (WaveNet, 2003).
Injection of private funding has also moved research and development of wave energy
converters towards their wider commercialisation with an increased number of potentially
attractive technological solutions.
Among the many types of wave energy converters and related technologies that have
been studied and have reached a pre-commercial stage of development, Oscillating Water
Column (OWC) Wave Energy Converters (WECs) have attracted particular attention.
The concept of an OWC has been subject to an extensive spectrum of research
activities worldwide. A variety of OWC technologies have been developed, which range
from small prototypes to full scale devices that have been built and tested over recent
years demonstrating the high level of technical maturity, which can satisfy the
requirements of industrial and commercial application.
However, despite all the achievements in commercialisation of OWC systems, there
are a number of issues which require further research and development. One of the most
important is how to increase the overall efficiency of existing and proposed OWC wave
energy converters and their components. The air turbine employed by any OWC system
is one of the major components that can significantly influence the overall system
efficiency.
In order to improve the actual efficiency of a chosen air turbine and to enhance the
overall OWC performance, it is important to fully understand the aerodynamic behaviour
of the turbine and to comprehend how the turbine interacts with other major OWC
components, such as the OWC chamber, nozzle-diffuser and nacelle.
2
Chapter 1
1.1 Motivation and objectives

Despite evident progress attained in conversion of oscillating and reversible
pneumatic power to electricity by means of self-rectifying air turbines (as covered in
Chapter 2) it is clear that there is still much to be done to increase their efficiency and to
extend their operational range and technical reliability.
1.1.1 Identification of research issues

On the basis of the literature review conducted during the present study several
important research problems have been identified.
The first problem concerns the methods, which are commonly adopted to estimate the
performance of OWC turbines. It was revealed that the most widespread method
providing quick and reasonably accurate performance estimation of the OWC turbines
based on the invention made by Dr. A.A Wells is the blade element momentum (BEM)
methodology combined with actuator disc principle. However, application of a blade
element/actuator disc analysis requires the input of reliable lift and drag data.
Unfortunately, this is not always attainable, especially when the designer needs to predict
the performance of turbine with novel blade shapes; such data-sets are simply not
available or even do not yet exist. To overcome this problem, the designer has to use an
appropriate tool to calculate the necessary aerodynamic coefficients.
The second problem is associated with the difficulty of applying the BEM method to
the analysis of the turbine rotors of relatively high solidity, which are common in OWC
air turbines. Lift and drag data that is usually used for BEM analysis is generally for
isolated aerofoils, yet in practical turbine rotors of high solidity there is “interference”
between adjacent blades or a “cascade” effect, which influences the isolated lift and drag
characteristics. However, the testing of aerofoils in a cascade to determine interference
factors is extremely difficult. Only a very few, and small, data-sets applicable to OWC
turbines are available in the public domain (e.g. Raghunathan, 1988). Most researchers
analysing the Wells turbine have used an analytical prediction of the interference factor
for a linear cascades based on the inviscid flow analysis developed by Weinig (1964).
However, there has previously been very little, if any, validation as to what extent the
Weinig interference factor data are correct for practical aerofoil cascades.
A third problem concerns the fact that the BEM method currently used for analysis of
the OWC turbines is a dimensional tool, which is applicable only for performance
3
Chapter 1
prediction of the turbine with specific dimensions. Universalisation of the BEM method
by employing a purely non-dimensional form will make it possible to analyse the range
of turbine geometries simultaneously.
The fourth research problem is mostly associated with limited data being available
regarding the effect of the major design parameters on the performance of OWC turbines
with variable geometry, viz. tip clearance, tc, blade thickness ratio, τ, blade aspect ratio,
AR, blade shape, hub-to-tip ratio, h, and rotor solidity, σ. Scarcity of CFD modelling of
these parameters can be also attached to the problem.
The fifth problem concerns the lack of published data on CFD simulations of the
entire pneumatic circuit of OWC wave energy plant, i.e. a full 3D model of the
pneumatic/aerodynamic system including turbine, nozzle-diffuser, OWC chamber,
nacelle, etc.
The sixth research problem relates to the fact that OWC air turbines are often
arranged to have a horizontal axis for convenience of construction. However, this forces
the airflow from the OWC chamber to make a 90° turn on its way to/from the air turbine
and can result in a non-uniform flow distribution at the inlet side of the turbine rotor.
From the viewpoint of the present author it is very important to look at the factors
causing this non-uniformity in detail as well as to investigate the conditions which
suppress this adverse flow distribution and provide more uniform flow field at the air
duct.
Finally, it should be mentioned that the issues regarding the technical reliability of the
turbines with variable geometry are beyond the research targets of the present thesis.
Further development of OWC WECs undoubtedly requires development of further
innovative technologies, which are needed to increase OWC efficiency and reduce the
cost of electricity production. A pertinent example of such innovations is the Denniss-
Auld turbine with variable rotor geometry, which has been developed by the company
Oceanlinx. Full scale tests of this novel turbine design at Port Kembla (Australia) have
shown that has potential to perform better over a wide operational range compared some
other the conventional OWC turbines. However, the development of this turbine has
raised a number of design issues, which are directly connected to some of research
problems identified above. This in turn, motivated the present author, to study these
problems by applying suitable analytical and numerical tools and gain new knowledge,
4
Chapter 1
which can be used further for optimisation and performance improvement of the OWC
turbines with variable geometry.
1.1.2 Primary objectives

Identification of the research questions in the previous sub-section helped to define
the framework of the present study and to formulate the prime objectives. Research
methods and tools employed for the present work included: the BEM analysis based on
the actuator disc principle, 2D CFD airflow analysis, full 3D CFD modelling of the
turbine in isolation and within a complete OWC system. For 2D cascade flow and 3D
CFD modelling of the flow through the turbine the ANSYS CFX-11 CFD package was
used. All CFD results regarding the turbine performance and other OWC components
were compared with available experimental data. A more detailed description of the
investigative methods is presented in Chapter 3.
On the basis of framework formulated above, the following specific objectives of the
present study were set as follows:
• Verification of the capability of ANSYS CFX to model isolated and cascade
aerofoils to obtain the aerodynamic coefficients necessary for BEM analysis.
• Determination of the cascade interference factor for the practical aerofoils used in
the OWC turbines by means of 2D CFD viscous flow simulations.
• Comparison of analytical interference factors based on the Weinig’s inviscid flow
analysis against the equivalents deduced from 2D CFD simulations.
• Formulation of a new non-dimensional multiple steamtube BEM model and
modelling of turbines with variable pitch blades.
• Prediction of performance of turbines with variable geometry using full 3D CFD
simulations.
• Determination of the effect of the major design parameters such as: tip clearance,
tc, hub-to-tip ratio, h, rotor solidity, σ, blade thickness ratio, τ, and blade shape on
performance of the full scale Denniss-Auld turbine model.
• Evaluation of various Denniss-Auld turbine configurations.
• Evaluation of pneumatic efficiency of the OWC system as a whole, viz. the OWC
chamber, nozzle-diffuser and the full scale Denniss-Auld turbine coupled to the
nacelle by means of 3D CFD modelling.
5
Chapter 1
• Investigation of the causes and effects of non-uniform flow in the nozzle-diffuser

upstream of the rotor during the exhalation cycle.
• Analysis of the experimental data obtained during the sea-trials of the Oceanlinx
OWC plant at Port Kembla and comparison with numerical results deduced from
the BEM and CFD simulations of the full scale Denniss-Auld turbine.
This thesis comprises ten chapters. The introduction, Chapter 1, provides an insight
into the subject matter and identifies the research issues along with formulation the
primary objectives of the project. An overview of the various OWC technologies, the
development of OWC turbines and introduction to CFD modelling of air turbines
designed to service OWC devices is given in Chapter 2. Chapter 3 provides a description
of the methodologies which were applied in the present study for aerodynamic analysis of
axial flow turbines. Chapter 4 then describes how the necessary aerodynamic coefficients
(lift and drag) were calculated using CFD analyses in order that the coefficients could be
used in the BEM analysis with confidence. The influence of “cascade effects” and the
determination of interference factors for linear cascades of staggered aerofoils are
described in the Chapter 5. Chapter 6 presents a new non-dimensional BEM model,
which was then applied to predict key performance characteristics of various OWC
turbines. In Chapter 7, results of three dimensional CFD modelling of OWC turbines, in
particular the Denniss-Auld design, are presented and discussed. Chapter 8 presents
results of 3D CFD modelling of OWC WEC systems comprising the major components
such as: chamber, nozzle-diffuser, and turbine with variable pitch blades coupled to the
nacelle in terms of defining its overall pneumatic efficiency. The factors influencing the
internal flow field in the horizontal air duct (nozzle-diffuser) are also investigated in this
chapter. Chapter 9 provides detailed analysis of the experimental data obtained during the
sea-trial tests of the full scale Denniss-Auld turbine and their comparison with numerical
results deduced from BEM and CFD analyses. The main body of the thesis closes with
conclusions and recommendations for future work in Chapter 10.
6
Chapter 2
CHAPTER 2 - DEVELOPMENT OF OWC TURBINES

In this chapter the basic concepts and configurations of OWC technologies are given.
Along with proven technological variants of the OWC WECs, the development of air
turbines designed to service these wave energy converters is reviewed. Both traditional
and modern methodologies for designing of OWC turbines are also discussed.
2.1 OWC technology as a practical means of wave energy conversion

Numerous publications and scientific reports acknowledge that OWC technology has
passed successfully from being demonstration models and prototypes to full-scale
applications. The three following sections provide a classification of OWC WECs
followed by a description of the working principles of the OWC concept and illustrations
of OWC technological variations.
2.1.1 Classification of OWC wave energy converters

Wave energy converters are generally classified according the orientation to the
incoming wave direction (Cruz, 2008).
• Terminators are placed with the principal direction at 90º to the direction the
incident waves, with beam much greater then length; and
• Attenuators are positioned perpendicularly to the incident wave crests and with
their beam much smaller than their length; and
• Point Absorbers are devices with no prevailing horizontal dimension and which
have a scale significantly smaller than the wavelength.
A schematic illustration of device orientation to the wave direction is given in Figure
2.1. This classification is defined with respect to the principal geometrical orientation of
the WEC to the wave direction. The principal geometrical orientation of the WEC must
be controlled in accordance to the current sea state; otherwise a lengthened converter can
be forced to act as a terminator or an attenuator at the same location under different sea
states. Correctly designed mooring configurations are required to control the WEC
orientation for particular sea states. A lengthened converter in attenuator mode usually
experiences significantly lower mooring forces than the same device in terminator mode.
When the WEC operates as a point absorber (e.g. OWC), it extracts energy from the
wave field over a region of no predominant orientation to the wave direction.
7
Chapter 2
ent
Incident waves Terminator Attenuator Point
Absorber
Figure 2.1 Schematic showing device orientation to the wave direction (Cruz, 2008).
The classification of WECs described above can be applied to the full spectrum of
existing and proposed wave energy converters. Cruz (2008) described another common
classification based on the device location:
• Onshore - the OWC wave energy converter is fixed on the shore; and
• Near-shore - the OWC plant is at some moderate distance from the shore; and
• Offshore – the OWC device is at a significant distance from the shoreline.
Literature devoted to extraction of wave energy defines onshore and near-shore OWC
devices as the First Generation Systems. These locations provide the simple and
economical way to build such devices using conventional construction techniques and
power take-off (PTO) equipment. Second Generation Systems include floating OWCs,
which are designed to operate at a wide variety of offshore and near-shore sites where
high levels of wave energy are available. Third Generation Systems are large-scale
offshore devices, both in terms of physical size and power output.
2.1.2 Working principles of OWC wave energy devices

The working principle of OWC WECs (Figure 2.2) is based on the oscillation of the
water column inside the OWC chamber, which drives a differential pressure across the
turbine rotor that, in turn, drives an electrical generator. The main components of the
system are the OWC chamber, the air duct, the air turbine and the electrical generator
located inside or external to the nacelle. The OWC chamber is the key part of the overall
8
Chapter 2
device structure with a number of important functions. Firstly, it couples the interior
hydrodynamic field with the incoming wave field, i.e. the incident waves are coupled to
the interior part of OWC chamber and cause the water column to oscillate. The lower part
of the submerged front wall of the OWC chamber should be designed to be below the
water surface at all times. Secondly, the OWC chamber couples the interior
hydrodynamic and aerodynamic fields via the interior water free surface, the rise and fall
of which drives air flow through the turbine. Finally, the OWC chamber prevents the
PTO mechanism from coming into direct contact with harsh environment of the sea water
and mitigates the forces arising during storm conditions.
OWC Valve
chamber Air turbine Nacelle
Free Air
Water
Surface
Incident
waves Air duct (nozzle-diffuser)
OWC
Shore
Sea-floor
Figure 2.2 A sketch of the cross-section of a generic shore-mounted OWC system.
For completeness, it must be noted that the interaction of the diverse processes within
the OWC system are complex. For instance, the turbine characteristics influence the
response of the OWC chamber, which in turn influences the hydrodynamic domain in the
vicinity of the device and vice versa. Some of the important issues regarding the
interconnection complexity of the hydrodynamic and aerodynamic domains as well as the
major design parameters influencing the performance of OWC systems will be
considered in the next section.
9
Chapter 2
2.1.3 Technological features of OWCs

There are a number of technological features that distinguish the OWC concept from
other technological principles of harnessing of ocean power. The first is associated with
the OWC chamber, which works as a “pneumatic” gearbox converting the slow internal
free surface motion to high-speed air motion through the air turbine. The conversion
efficiency of the OWC chamber can be characterised by the capture efficiency, which is
the ratio of the pneumatic energy to hydrodynamic wave input energy, where the input
energy is the product of the device width and the incident energy per metre of coastline
(Curran et al., 1998a).
As a result of the direct exposure of the OWC to wave forces of high magnitude,
many chambers have been built using reinforced concrete. This construction technique is
usually employed in shore-mounted OWC systems. However, there is a range of OWC
devices (e.g. floating) where the whole structure including the capture chamber has been
made of steel applying normal shipbuilding techniques. Most existing OWC plants are
equipped with chambers with vertical water columns as illustrated schematically in
Figure 2.2. However, some OWC systems use chambers with inclined water columns as
depicted in Figure 2.3. An example of such OWC chamber design is the LIMPET (Land
Installed Marine Power Energy Transmitter), which is a full-scale shore-mounted OWC
plant located on the Isle of Islay, Scotland (LIMPET, 2002). There are two distinct
advantages of inclined OWC chambers that improve their capture efficiency. Firstly, the
inclined column reduces entrance turbulence and internal sloshing. This is mainly true at
the seashore where shallow water effects increase the surge motions relative to heave.
Secondly, the inclination of the OWC increases its water plane area for a given chamber
cross section area. This in turn, allows the primary water column resonance, which is to a
large extent determined by the ratio of the OWC plane area to the entry area, to be
coupled to the major period of the incoming waves (Heath et al., 2000).
The important aspects of any OWC chamber are the configuration and thickness of
the front wall that must guarantee device survivability under extreme sea conditions,
particularly from the forces associated with wave slamming.
10
Chapter 2
Figure 2.3 Inclined OWC chamber of LIMPET (LIMPET, 2002).
Another approach to increase capture efficiency of an OWC system has been

implemented by Oceanlinx previously known as Energetech Australia Pty. Ltd.
Oceanlinx has developed a technology that employs a large parabolic-shaped collector
for a full-scale sea-bed mounted OWC device (Figure 2.4) known as their Mark 1
technology. According to Finnigan (2004), a parabolic-shaped wall amplifies the wave
height at the focal point by a factor of approximately 2.5 or more, depending on period.
Such focusing action increases the production capacity of the OWC device. However, the
actual shape of the parabolic wall affects its capacity to amplify the wave height and must
be optimised prior to its construction and installation. It was found that the moderately
flattened parabola optimally amplifies the waves and does not significantly increase its
construction cost. Besides the shape issue, it was also found that the installation of
parabolic collector requires provision of a gap between the wall and sea-bed. However,
the gap causes energy losses and the gap size must be optimised in order to minimise
them. Tests to clarify this aspect revealed that a gap of 4m between the sea-bed and the
bottom of the collector did not result in appreciable energy loss in conditions where the
total water depth was 10m or more (Finnigan, 2004). Despite the difficulties of
construction and installation, the parabolic-shaped collector has an additional advantage
over conventional OWC systems; the parabolic wall allows efficient conversion from
incoming waves over a range of angles.
11
Chapter 2
Figure 2.4 Top view of Oceanlinx OWC device (Finnigan, 2004).
The second technological feature that distinguishes the OWC concept from
alternative wave energy conversion methods is the design of the PTO, which is usually an
axial flow air turbine. The oscillating and reversible nature of the airflow through the
turbine is a major and unique design challenge which is not encountered elsewhere in the
wide range of axial flow turbine applications. In order to overcome this challenge, the
majority of OWC WECs are equipped with self-rectifying air turbines, which are able to
maintain unidirectional rotation of the rotor in reversible airflow conditions.
Today the most common turbine type for OWC applications is the monoplane Wells
turbine with fixed pitch blades invented by Dr. A. A. Wells in 1976 (Raghunathan et al.,
1985). A schematic view of this turbine, which commonly consists of several
symmetrical aerofoil blades staggered at 0° relative to the plane of rotation around a hub,
is presented in Figure 2.5.
12
Chapter 2
Rotor
Rotor
Air
Air
Figure 2.5 Schematic of the Wells turbine.
The basic principle of the unidirectional rotation of the Wells turbine lies in the use of
aerofoil blades symmetrical about the chord line and in the action of forces on them as
shown in Figure 2.6. The relative airflow velocity, WR, generated by the combination of
the axial flow velocity across the turbine, VA, and the tangential velocity of the blade, VT,
creates aerodynamic forces on the aerofoil section depending on magnitude of the angle
of incidence α. These forces are a lift force, FL, and a drag force, FD, acting normal and
parallel to WR respectively. The forces FL and FD can be resolved into the tangential and
axial forces FT and FA respectively as:
FT = FL sin α − FD cos α
F A = FL cos α + FD sin α
For an aerofoil operating in an oscillating and reversible airflow, the magnitudes and
directions of FT and FA vary during a wave cycle. However, the direction of FT remains
unchanged giving the Wells turbine its self-rectifying property.
13
Chapter 2
FL
FA
WR
VA
α
FD
VT
FT
VT FD
VA α
WR FA
FL
Figure 2.6 Aerodynamic forces acting on a blade of a Wells turbine rotor.
Another self-rectifying turbine type that has been employed in some full-scale OWC
projects is the impulse turbine with guide vanes (GVs) (e.g. Kim et al., (1988)). A
schematic of the impulse turbine rotor with fixed GVs is illustrated in Figure 2.7. The
working principle of the impulse turbine is based on the change of the velocity of the air
jet which impinges on the turbine's cup-shaped blades altering the airflow direction. The
change in momentum (impulse) direction provides the turbine rotation without the
occurrence of a significant pressure drop across the rotor. The self-rectifying capability of
an impulse turbine is attained by means of symmetrical guide vanes fixed at setting angle,
θ, which act as a nozzle/diffuser pair in reversible airflow.
An example of a relatively recently designed and built full-scale turbines for OWC
applications is the Denniss-Auld variable pitch angles turbine (Figure 2.8) installed in
Oceanlinx OWC plant at Port Kembla, Australia. This turbine is not modelled on either
the Wells or the impulse turbine configurations, and is different from both in many
respects (Alcorn and Finnigan, 2004). It can be clearly seen that the turbine has specially
shaped blades which are symmetric about the mid-chord plane. An actuation mechanism
is used to vary the pitch of the blades so that as the airflow changes direction the turbine
blades swap orientation maintaining a unidirectional rotor rotation (Figure 2.9).
14
Chapter 2
GV GV
θ
Rotation
Rotor
GV GV
Flow
Figure 2.7 Schematic of an impulse turbine rotor with fixed GVs.
Figure 2.8 A full-scale Denniss-Auld turbine with variable pitch angle blades prior to
installation (Alcorn and Finnigan, 2004).
15
Chapter 2
Figure 2.9 The turbine pitching regime of the Denniss-Auld turbine (Alcorn and
Finnigan, 2004).
In order to increase the pneumatic efficiency of the OWC plant, the turbine is located
at the mid-section of a convergent-divergent duct (see Figure 2.10) which accelerates the
airflow through the turbine to maximise turbine efficiency. It also can be seen that the air
duct is arranged horizontally at 90° to the main-flow in the OWC chamber. This duct
arrangement, in contrast to the vertically oriented configuration facilitates access to the
turbine and electrical generator components for maintenance and repair, but can have
adverse effects as discussed later in Chapter 8.
Figure 2.10 Power train schematic (Alcorn and Finnigan, 2004).
An important practical feature shown in Figure 2.2 is the shut off valve, which is
designed to protect the turbine during large wave excursions. There are a number of the
valve configurations employed by existing and proposed OWC WECs. By-pass pressure-
relief valves are commonly incorporated in the roof of the OWC chamber and throttle
valves are installed in the series with the air turbine. The description of the technical
16
Chapter 2
facets and design variations of the valves are beyond the scope of the present study.
However, detailed information regarding their influence on performance of the OWC
system can be found in publications such as that by Falcao and Justino (1999).
2.1.4 Design variations of OWCs

In this section, design variations of OWC projects that have been developed to full-
scale are presented in accordance to the following classes:
• Shore-mounted systems;
• Sea-bed devices;
• Floating systems.
2.1.4.1 Shore-mounted OWC systems

The literature review conducted under the present study revealed, that the majority of
previous OWC WECs have been shore-mounted. The two best known full-scale OWC
shore-mounted systems are the LIMPET plant on the Islands of Islay in Scotland and the
Pico plant on one of the islands of the Azores.
A schematic view of LIMPET is shown in Figure 2.3. This 500kW shoreline OWC
wave power system was built between 1998 and 2001 and commissioned during the
spring of 2001 to replace a 75kW prototype constructed ten years earlier and located on
an adjacent site (LIMPET, 2002). Operational experience and experimental data gained
during exploitation of the 75kW prototype provided valuable information used to design
and build a new LIMPET station. The power take-off mechanism of the LIMPET was
2.6m diameter contra-rotating biplane Wells turbine. Each turbine rotor was mounted
directly onto the shaft of an induction machine and was balanced with a flywheel on the
other end of the shaft. A single turbine rotor was rated at 250kW giving the combined
biplane design output of 0.5MW. In addition, two valves (a butterfly and a vane) were
included between the rotors and the OWC chamber.
The major outcomes of the LIMPET project can be articulated as follows. First, it
represented an important step in the development of wave power technology
demonstrating its commercial potential. Secondly, it played an important role in the
development of the next generation of OWC systems as well as other device types.
The pilot plant on the island of Pico in the Azores (Portugal) is another full-scale
shore-mounted OWC system (400kW), which was built in 1995-1999, with funding from
the European Union. The 12m × 12m concrete structure of the OWC chamber was
17
Chapter 2
constructed in a small natural harbour (in a water depth of about 8m). A horizontal-axis
Wells turbine-generator set was used as PTO equipment in the original plant. The
induction type generator was capable to operate over a relatively wide range of speeds.
The automated electrical generation system of the OWC plant was able to supply
electricity to the island grid on a permanent basis (Falcao, 2000). After a brief test run in
1999 the electrical and control equipment were damaged and further tests could not be
carried out. The Pico project was suspended for several years. The plant was revived by
the Wave Energy Centre (Portugal), with support from a new Portuguese funding scheme
for scientific pilot projects (PRIME/DEMTEC) in 2003-2005.
Figure 2.11 Scheme of the OWC Pico plant and its components (Neumann et al., 2007).
The refurbished Pico plant including its components is shown schematically in Figure
2.11. A horizontal axis Wells turbine with a rotational speed in the range of 750-1500
RPM and an asynchronous generator of 400kW were used for the new project. Also
during the recovery works a relatively slow by-pass relief valve was installed on the top
of the OWC chamber of the plant. The main function of the valve was to allow energy
dissipation to the atmosphere in very energetic sea states. In order to improve the
aerodynamic performance of the Wells turbine, the guide vane stators were installed
upstream and downstream the rotor. To isolate the turbo-generator set from the OWC
chamber and thus enhance safety, two different valves were installed in the turbine duct:
a fast acting butterfly valve close to the turbine and a slow “gate-valve” next to the
entrance of the duct.
18
Chapter 2
After the preliminary testing phase of Pico plant some general project conclusions
have been drawn:
1. Despite the long period of inactivity, the refurbished Pico plant successfully
commenced its operation and delivered electricity to the grid.
2. The project achieved its primary objective of demonstrating the capacity to contribute
to a local grid supply over a representative period of time.
3. Very valuable operational experience was gained from this initial testing phase.
2.1.4.2 Sea-bed mounted OWC devices

The largest of all sea-bed mounted OWC devices to date was known as the OSPREY
(Ocean Swell Powered Renewable Energy) built in 1996 in Glasgow (Figure 2.12). The
OSPREY project was developed and coordinated by Applied Research and Technology
Ltd., (WaveNet, 2003). The device was designed to operate in moderate water depths (i.e.
< 20m) at the site of Dounreay off the north coast of Scotland. The whole structure of
OSPREY was made of steel and for its construction normal shipbuilding techniques were
applied. For electrical production, the device was equipped with a Wells turbine coupled
to the generator.
The first OSPREY was rated as a 2MW wave energy converter with the possibility of
installation a 1.5MW wind turbine at a later date. Combination of OWC device with a
wind turbine was considered as an additional advantage to maximise overall power
output and minimise capital cost, as the wind turbine utilizing the near shore winds can
be mounted on top of the wave energy converter. Unfortunately, irrevocable damage to
the first OSPREY occurred during its deployment, so no performance data exist to
determine the device efficiency. The design of a new OSPREY 2000 was based on the
composite construction. Besides that the installation procedures were planned to
minimise the time required to deploy the OSPREY 2000 in open waters. According to
Clement et al., (2002), the new device should operate in 15m of water within 1km of the
shore, generating up to 2MW of power for coastal consumers. Unfortunately, the
OSPREY 2000 was never built.
19
Chapter 2
Figure 2.12 OSPREY during float out from the John Brown shipyard, Glasgow, Scotland
(WaveNet, 2003).
It has been known theoretically and experimentally since the early 1980s, that the
absorption of wave energy can be increased by physically extending the capture width of
the WECs by using the wave collector based on natural or man-made walls in the
direction of the waves (WaveNet, 2003). This idea has been put into practice in a number
OWC prototypes.
An Australian company Oceanlinx has incorporated a large parabolic-shaped
collector into a full-scale sea-bed mounted OWC device (Figure 2.13). The large size of
the 40m wide parabolic wall as compared to the dimensions of the 10m wide OWC
chamber itself presented earlier in Figure 2.4 was one of the main novelties of the full-
scale prototype (OCEANLINX, 2009). Another novel feature was the self-rectifying
variable-pitch turbine (see Figure 2.8). Oceanlinx Mk1 was fitted out and first deployed
at Port Kembla, 75km South East of Sydney in 2005. Its operation over the last few
years had provided valuable operational data, which has helped to guide the development
of subsequent designs. The Mk1 prototype has completed its function at Port Kembla and
has been decommissioned.
20
Chapter 2
Figure 2.13 Side view of sea-bottom mounted Oceanlinx OWC device (Finnigan, 2004).
2.1.4.3 Floating OWC plants

Floating OWC projects began to draw attention in the late 70s, after the Japan Marine
Science and Technology Centre (JAMSTEC) tested the world's first large-scale offshore
floating prototype KAIMEI shown in Figure 2.14. The KAIMEI was 80m long and 12m
wide. The device had 9 generators, which were mounted above lengthwise OWC
chambers open to the sea at the bottom. Action of incoming waves caused the internal
water levels in chambers to move up and down, forcing reversible airflow to drive the air
turbines. Open sea tests of KAIMEI were carried out from August 1978 to March 1986,
three kilometres offshore of Yura, Tsuruoka city, Yamagata Prefecture, Japan
(JAMSTEC, 2009). Development of floating OWC prototypes in Japan has been
continued e.g. the "Mighty Whale" offshore OWC plant (Figure 2.15) was designed, built
and tested by JAMSTEC. The prototype name of "Mighty Whale" was given since the
plant resembled a whale in appearance, and was 50m × 30m × 12m in overall
dimensions. It was designed to float at a draft of 8m at even keel, and its mooring system
was made to withstand wind-wave conditions resulting from a 50 year storm (JAMSTEC,
2009).
21
Chapter 2
Figure 2.14 The world's first large-scale offshore floating prototype KAIMEI
(JAMSTEC, 2009).
Figure 2.15 The OWC plant "Mighty Whale" (JAMSTEC, 2009).
The "Mighty Whale" had three OWC chambers distributed in a row and each
chamber was equipped with a 1.7m diameter, biplane self-rectifying Wells turbine made
of aluminum-alloy. Each turbine rotor had 8 blades with the NACA0021 profile. The
overall rated power capacity of the device was set at 120kW. The "Mighty Whale" began
operating in 1998 in Gokasho Bay, Japan at 40m depth providing valuable experimental
data and operational experience until its decommissioning.
22
Chapter 2
A major conclusion made by JAMSTEC after completion of “Mighty Whale” testing

was that offshore OWC floating devices have the following advantages over shore/sea-
bed mounted devices. First, substantial economies result from the fact that floating OWC
devices experience significantly lower impact loads in extreme sea states. Secondly, the
increase of water depth generally increases available wave energy (exceptions are
shallow-water regions with favorable bottom topography, which can lead to localized
focussing of wave energy). Thirdly, the rigid-body motion of the floating device relative
the motion of the OWC can increase absorption of energy in certain wave conditions.
One of the recent floating OWC projects is 1/3rd scale prototype Mk2 (Figure 2.16)
designed and built by Oceanlinx Australia. The purpose of the Mk2 is to obtain detailed
technical data for development of full-scale floating OWC plants (Oceanlinx, 2010).
Figure 2.16 Oceanlinx floating OWC 1/3rd scale prototype Mk2 (Oceanlinx, 2010).
Another recent floating OWC project is OEBuoy developed by OceanEnergy Ltd

(Ireland). The OEBuoy is a wave power generation platform which the company claim
was designed for survivability (Figure 2.17). The OEBuoy pilot plant was successfully
deployed on the west coast of Ireland in 2006 (OceanEnergy, 2011). The details of a
novel impulse turbine configuration that is used in OEBuoy project can be found in work
by Thiebaut et al., (2011).
23
Chapter 2
Figure 2.17 OEBuoy (OceanEnergy, 2011)
The wave energy converters presented above reflect a spectrum of all technical
solutions based on the OWC concept, but of course not a completed list of the full-scale
devices developed over past years. In order to avoid unnecessary repetition of the design
variants already described, other full-scale OWC systems have not been presented here.
2.2 Design requirements for optimal operation of OWC systems

In order to provide an input to design considerations, it is important to evaluate the
OWC system performance through the range of expected operating conditions. In this
section the key design parameters influencing performance of OWC systems are
described. After a short introduction to geometrical and control parameters, which can be
used for performance optimisation of OWC devices, the first sub-section gives an insight
into performance modelling of OWC articulating the importance of proper coupling of
the hydrodynamic and aerodynamic subsystems. The next sub-section describes
parameters influencing performance of the OWC system. The third sub-section considers
the factors affecting the applied damping required to be developed by OWC turbines.
According to Cruz (2008) any wave energy converter can be considered as a
mechanical system having both geometrical and control parameters. Geometrical
parameters define the structure of the WEC and cannot be changed once the device has
been built. Control parameters are commonly variable parameters, capable of being
adjusted via a PTO mechanism to match the wider sea-state spectrum.
24
Chapter 2
2.2.1 Modelling of the OWC system

Several approaches have been employed in the past to model and optimise OWC
systems. The simplest theoretical model assumes the incident waves to be regular or
monochromatic. This approach allows the analysis to be performed in the frequency
domain provided the air turbine characteristic is approximately linear (Evans, 1982); in
this case the hydrodynamic coefficients of radiation and diffraction can be obtained, for
instance, analytically for simple geometries (Evans, 1982), as well as numerically (Lee et
al., 1996) or experimentally (Sarmento, 1993). In contrast to frequency domain
modelling, time-domain method can provide more realistic analysis irregular waves and
with non-linear power take-off equipment. This requires the knowledge of the
hydrodynamic coefficients as functions of frequency and more computational resources.
However, it yields performance variables, including power output, as functions of time
(Falcao and Justino, 1999).
However, existing models tend to describe the hydrodynamics and aerodynamics of
the OWC device separately. In the context of optimisation it appears crucially important
to be able to deal with complete system models rather than with separate models of
hydrodynamic and aerodynamic domains.
2.2.1.1 Coupling of the hydrodynamic and aerodynamic domains

Among the earliest works devoted to the mechanisms of hydrodynamic-aerodynamic
coupling is the work reported by Evans (1982). Evans provided a general theory for any
number of fixed rigid bodies each enclosing an internal free surface experiencing an
increased pressure relative to the atmosphere. He also demonstrated that there was a
parallel with the rigid-body motion, provided the ideas of exciting forces on fixed bodies
and the given velocities of moving bodies were replaced by the ideas of volume flux
across the internal free surfaces open to the atmosphere and the given forcing pressures
on those free surfaces, respectively.
Interesting results were reported by Sarmento et al. (1990). These authors used linear
wave hydrodynamics to derive the exact theoretical conditions required for a specified
PTO to extract the maximum energy from regular or irregular waves. In another work
(Evans et al., 1995) a two-dimensional shore-mounted OWC system was optimised with
regard to its hydrodynamic efficiency in terms of the design parameters of submergence
of the barrier, length of the OWC chamber and turbine constant. This work was extended
by Weber (1999) and Weber and Thomas (2000) with the integration of air
25
Chapter 2
compressibility effects and they showed significant compressibility impacts on the

optimised design parameter values and on the performance of the OWC device during
operation.
In the paper by Weber and Thomas (2001), a numerical model based on the
incorporation of a Wells turbine into the power conversion chain was presented
demonstrating the importance of the OWC chamber design with respect to its principal
dimensions in the overall optimisation of the power conversion properties.
It is also important to mention that Weber and Thomas paper “Optimisation of the
hydrodynamic-aerodynamic coupling for an Oscillating Water Column wave energy
device” (Weber and Thomas, 2000) identified the major hydrodynamic-aerodynamic
issues and represented the OWC system as consisting of number different physical sub-
domains, which must be coupled accurately when the OWC system is modelled. It is seen
from Figure 2.18 that interconnecting quantities on conjoint boundaries of neighbouring
domains reflect the flow of information through the overall OWC system. In particular,
the modelling of the interior air flow problem plays a major role in both aspects of the
hydrodynamic-aerodynamic coupling, namely the linkage of the OWC to the turbine and
the coupling of the diffraction and radiation problems in linear water wave theory.
Figure 2.18 Domains and their problems, extracted from (Weber and Thomas, 2001).
Thus, the main aim of accurate coupling of the hydrodynamic and aerodynamic
domains is to provide more realistic algorithms for OWC system optimisation (Weber
26
Chapter 2
and Thomas, 2000). A detailed description of the completed algorithm of the proper
hydrodynamic-aerodynamic system coupling, applicable to optimisation of given OWC
structure is beyond the scope of the current sub-section. However one of the essential
subsystems coupling mechanisms i.e. the interconnection between the motion free-
surface of the water and the turbine is now described as it is direct relevance to the
present work.
As a result of the incident wave action, the free-surface motion inside the OWC
chamber displaces a volume flow rate of air, q(t), and produces an oscillating air pressure
P(t) + Pa, where Pa is atmospheric pressure. Using liner wave theory (Evans, 1982), the
volume flow rate, q(t), can be expressed as the sum of the diffraction air flow rate, qd(t),
due to the incident waves when the internal pressure is kept constant and equal to Pa, and
the radiation flow rate, qr(t), caused by the oscillating air pressure in the OWC chamber
in the absence of incident waves:
q(t) = qd(t) + qr(t).
The mass flow rate of air leaving the chamber through the turbine is given as:
d (ρ aV )
m& = − , (2.1)
dt
where ρa and V are the density of air and the volume of air inside the chamber,
respectively. Assuming that the relative variations in ρa and V are small, and that ρa is
related to the pressure P + Pa through the linearised isentropic relationship (Falcao and
Justino, 1999). With dV/dt = - q, the mass flow rate m& can be defined as:
V 0 dP
m& = ρ a q − , (2.2)
ca2 dt
where ρa and ca are the density of air and the speed of sound in atmospheric condition
respectively, and V0 is the undisturbed value of V.
Turbomachinery theory states that the air mass flow rate through the turbine and
generated aerodynamic torque depend on the pressure p and on the rotational speed ω.
This theory can be applied to relate m& to P. In order to evaluate these quantities it will be
necessary to introduce the performance characteristics of the turbine. Ignoring the effects
of varying Reynolds number and Mach number, these characteristics can be written in
dimensionless form as (Dixon, 1978).
Π = f P (Ψ ) (2.3a)
Φ = f Q (Ψ ) , (2.3b)
27
Chapter 2
where
P m& Wt
Ψ= 2
, Φ= , Π= . (2.4)
2
ρ a ω DT ρ a ω DT 3
ρ a ω 3 DT 5
Here Ψ is non-dimensional pressure (or head); Φ is non-dimensional flow rate; Π is non-
dimensional turbine power output; ω, rotational speed (rad/s); Wt, turbine power output
(mechanical losses are ignored), and DT, turbine rotor diameter (or rotor tip diameter).
2.2.1.2 Parameters influencing performance of the OWC

In context of the present sub-section a simple model illustrating the parameters
influencing OWC performance is now presented, whereby the OWC is modelled by the
well-known equation describing the damped motion of a body oscillating with a single
degree of freedom due to a time varying force (Curran et al., 1997b):
d2y dy
F (t ) = m 2
+B + Ky , (2.5)
dt dt
where F is the applied force at time t, m is the mass of the body, B is the damping, y is the
displacement and K is the spring constant. The spring restoring constant K is defined as:
K = ACρwg, where AC is the free surface area of the OWC, ρw is the density of water and g
is the coefficient of gravitational acceleration. A direct analogy can be drawn with a
mechanical mass-spring damper system, although two additions must be made in order to
account for the waves generated by the OWC as it oscillates. Firstly, the mass of the
OWC is said to comprise two components, including the frequency-dependent added
mass. For convenience the total mass of the OWC water column can be expressed as ME.
Secondly, the system damping also consists of two components, namely: the secondary
damping B2 due to wave radiation and losses, and the applied damping BA. The applied
damping extracts energy from the OWC and is provided by the turbine. A solution to the
equation of motion for sinusoidal excitation was proposed by Mei (1976).
2 2
B AωOWC F
WOWC =
(( 2
2 K − M EωOWC ) + (B
2
A
2
+ B2 ) ωOWC
2
), (2.6)
where WOWC is the average pneumatic power output in watts and ωOWC is the angular
frequency of oscillation in rad/s.
In order to maximise the power output of the OWC an optimum value of turbine
applied damping may be deduced as:
28
Chapter 2
2
 K − M EωOWC
2

B AOPT = B22 +   . (2.7)
 ωOWC 
2
It is seen, that resonance of the OWC occurs when K = M E ωOWC . Thus, the optimal value
of applied damping at the resonant frequency becomes equal to the secondary damping. It
is also evident, that the optimal level of applied damping increases with increasing
frequency. It should be noted from Eq. 2.7 that no single level of damping will optimise
the OWC performance through all expected sea-states. This statement given above is
supported by the results published by Curran et al., (1997b), who reported the results of a
series of hydraulic model tests to examine the influence of applied damping on an
OWC’s power output and bandwidth response (Figures 2.19 and 2.20). It can be seen
from Figure 2.19 that the power output of the column, WOWC, increased rapidly with
applied damping, BA, to a maximum value, after which it decreased steadily. The
performance of the OWC was thus very sensitive to variations in BA at levels which are
lower than that of its optimum. Therefore, if the OWC system is under-damped this leads
to considerable efficiency reduction in contrast to the over-damping, which gives a
stimulus to over-damp the system design.
Figure 2.19 Influence of pneumatic damping and wave height on power output of the
OWC (Curran et al., 1997b).
29
Chapter 2
Figure 2.20 Influence of damping and frequency ratio on the capture factor, CF, of OWC
(Curran et al., 1997b).
The performance bandwidth for different damping levels, at an initial value of BA1,
and for double (×2), and quadruple (×4) the initial level is shown in Figure 2.20. The
capture factor, CF, is defined as the ratio of the capture width of the device to the actual
OWC width (Curran et al., 1997b), and is plotted as a function of the frequency ratio
f=ωOWC/ωO, where ωO is the resonant frequency. It is evident that in order to maximise
the output of the OWC device for any particular sea-state its performance must be
optimised across a frequency spectrum.
It is clearly seen from these examples that optimum pneumatic power is only
produced in a wide range of the sea-states if the applied damping, BA, can be adjusted
appropriately. Thus, the applied damping, BA, provided by the air turbine is a key design
characteristic influencing the performance of the OWC.
2.2.1.3 Parameters influencing applied damping of OWC turbines

According to Curran et al., (1997b), the damping, BA, applied to the OWC by the
Wells turbine can be expressed in terms of the total pressure drop, ∆P0, the cross-
sectional area of the water column surface, AC, and the air velocity in the OWC chamber,
VC, as follows:
AC A2
B A = ∆P0 = ∆P0 C . (2.8)
VC AAV A
Eq. (2.8) can be also expressed introducing the damping ratio BR = P*/φ, to:
30
Chapter 2
 A2 C   P* 
B A = 4 ρ a  VT   . (2.9)
 AA  φ 
In turn, Eq. (2.9) can be further rearranged giving:
1  P* 
BA = ( )
πρ a ΩDT3 AR2 1 − h 2 VT   . (2.20)
2 φ 
Here the tip diameter is given by DT = 2VT/Ω (Ω is the rotational speed of the rotor), the
1
column-duct ratio by AR = AC/AA, the turbine duct area by AA =
4
( )
πDT2 1 − h 2 and the
hub-to-tip ratio by h = DH/DT, where DH is the hub diameter.

The damping ratio, BR, can be expressed in terms of plane number, Np, and solidity, σ,
as shown in Figure 2.20. The solidity being a ratio of the total blade area in plane view,
AB, and turbine annular duct area, AA, is given as: σ = AB/ AA. The results presented in
Figure 2.20 are small-scale and full-scale data obtained by Curran et al., (1998a) in the
laboratory and at the Islay prototype plant, respectively. Empirical results by Eves (1986)
are also included in Figure 2.21 for comparison with lab-scale and full-scale data.
The solid curve on Figure 2.21 illustrating relationship between BR and σ expressed
as:
π 
B R = 0.525 N p tan  σ  . (2.21)
2 
The value of 0.525 is a correlation coefficient, which was introduced by Curran et al.,
(1997b) in order to take into account the scale effects due to the variation in Reynolds
number. Finally, substituting Eq. (2.21) into Eq. (2.20) gives:
π 
( )
BA = 0.2625 ρ a ΩDT3 AR2 N p 1 − h 2 tan σ  . (2.22)
2 
It is seen from Eq. (2.22) that applied damping, BA, is a function of the following
design parameters: Ω, DT, AR, h, Np, σ, though two of them DT and AR are interrelated.
This is also applicable for other OWC turbines presented in the next section.
31
Chapter 2
Figure 2.21 Solidity effect on damping ratio of Wells turbine (Curran et al., 1998a).
However, it should be noted that the discussion above is primarily relevant to PTOs
with linear pressure vs volume flow characteristics (e.g. Wells turbine). In the case of
non-linear OWC PTOs the question of optimal pneumatic damping is significantly more
complex, optimum pneumatic damping is not independent of wave height (see Sarmento
& Falcao (1985) for example).
2.3 Development of air turbines for OWC applications

The literature review during the present study showed that air turbines being currently
used or proposed for OWC applications can be divided on two major groups:
• Turbines with fixed pitch blades;
• Turbines with variable pitch angle blades.
The first group includes the conventional Wells turbine and some modifications, e.g.
with inlet and outlet fixed guide vanes, biplane designs and contra-rotating rotors.
Technological simplicity and robustness in addition to the capability of self-rectification
are the main advantages of this turbine group. Axial and radial impulse turbines with
fixed guide vanes can be also added to the first group. Despite increased design
complexity, impulse turbines are able to operate over a wider range of flow conditions
compared to the turbine types based on the Wells turbine configuration.
The second group includes air flow turbines with variable rotor geometry. These
turbines can be listed in accordance with increasing design complexity as follows: the
Wells turbine with self-pitch controlled blades: the Wells turbine with variable pitch
32
Chapter 2
angle blades; and the Denniss-Auld turbine with variable pitch angle blades. The variable
rotor geometry in contrast to the fixed blades imposes an extra complexity in design of a
turbine, but may significantly improve its efficiency over a much wider range of airflows
generated by the OWC.
2.3.1 Air turbines with fixed pitch blades
2.3.1.1 Monoplane Wells turbines

The Wells turbine has been the most common type of air turbine with fixed pitch
blades employed by OWC wave energy devices. Besides a self-rectifying capability the
Wells turbine has some other strong facets such as: technical simplicity, reliability and
design robustness. However, these advantages are offset by its limited operational range.
The aerodynamic performance of the Wells turbine is commonly described by the
following dimensionless coefficients (Curran et al., 1997b):
*
W ∆ P0 VA V W P
W =*
3 5
, P* = 2 2
,φ = = A , η= , BR = .
ρ a Ω DT ρ a Ω DT VT Ω RT ∆P0 Q φ
Two of them W* and P* were introduced earlier as Π and Ψ presenting the non-
dimensional values of power output and pressure accordingly. The variables W, ∆P0, ρa,
Ω and DT present the power output, the pressure drop across the turbine, the density of
air, the rotational speed of the rotor, and the rotor tip diameter, respectively. Note: the
turbine tip radius, RT, may also be used for non-dimensional coefficients. The variable φ
is the flow coefficient expressed by ratio of the axial air velocity, VA, and tangential
velocity of the blade tip, VT. In turn, VA = Q/AA, where Q and AA are the axial flow rate
and the annular duct area at turbine location, respectively. The flow coefficient, φ, can be
also expressed as φ = tan (α), where α is the angle of incident shown in Figure 2.6. The
final characteristics summarising the turbine performance are efficiency, η, and the
damping ratio, BR. Typical variation of η and P* with φ for the monoplane (MP) and the
biplane (BP) Wells turbine configurations are shown in Figure 2.22.
33
Chapter 2
φ
Figure 2.22 Performance characteristics of the monoplane and the biplane (Curran and
Gato, 1997a).
It is seen from Figure 2.22 that the aerodynamic efficiency of the monoplane Wells
turbine, η, increases with flow coefficient, φ, up to a certain value then it decreases due to
the change in the angle of incidence, α, of airflow onto the blades (see Figure 2.6). At
higher values of angle of incidence the boundary layers on the aerofoil blades tends to
separate, thereby increasing the drag force, FD, and reducing the lift force, FL. The
phenomenon of flow separation (see Figure 2.23), known as turbine stall, can be severe
and even leads in extreme cases to the negative turbine torque. However, it has been
found that flow separation occurs more gradually on thick aerofoils resulting in a more
gradual drop of η with increasing φ. According to Raghunathan (1995) the Wells turbine
having NACA aerofoil profiles of 20% thickness demonstrated the best performance
compared to thinner aerofoils from the same series. It also was found that thick airfoils
provide better self starting characteristics in the Wells turbine. Setaguchi et al., (2003a)
studied the influence of the blade profile on starting and running characteristics under
sinusoidal flow conditions and confirmed the same effect of the aerofoil thickness ratio, τ,
which is the ratio of the maximum thickness of an aerofoil section to the length of its
chord, on performance of the Wells turbine as Raghunathan (1995). This highlights the
importance of the blade profile geometry and in particular the aerofoil thickness as key
design parameters.
34
Chapter 2
Direction of rotation
VT α
WR
W
VA
Figure 2.23 Visualisation of flow separation by means of CFD (from the present study).
2.3.1.2 Key design parameters affecting the Wells turbine performance

Besides the aerofoil thickness ratio, τ, there are other design parameters affecting
performance of the Wells turbine in terms of its starting and running characteristics. In
this sub-section a short description of other turbine design parameters presented in Figure
2.24 will be given.
As defined earlier, the turbine solidity is the ratio of the total blade area in plane view,
AB, to the turbine annular duct area, AA: σ = AB/AA. This important design parameter is
also known as a measure of blockage to airflow through the turbine and as a turbine
characteristic, which determines the mutual interference between blades.
The solidity can also be expressed as:
Nc Nc
σ= = , (2.23)
π (RT + RH ) πRT (1 + h )
where, N is the number of blades; c is the chord length of the blade, RT, is the tip radius,
RH, is the hub radius and h is the hub-to-tip ratio.
35
Chapter 2
Guide
vanes
Number
Rotor of planes
diameter
Aerofoil
thickness
Aspect ratio
ratio
Performance indicators
•Power output
Wells •Efficiency
Turbine
solidity
turbine •Pressure drop
•Flow rate
•Operational range
Hub to
tip ratio
Generator
characteristics
Tip Rotational
clearance speed Inlet turbulence
M and Re
Figure 2.24 Design parameters affecting performance of the Wells turbine.
According to Raghunathan (1995), Wells turbine efficiency is not very sensitive to

variations in solidity up to a value of 0.5, but a decrease in efficiency is noticeable for σ >
0.5. Efficiency reduction at high solidity is said to be generally a result of increased
losses of kinetic energy at the turbine exit associated with swirl. Boundary layer
separation due to interaction between the blades and the hub as well as 3D effects in the
vicinity of the hub may also lead to decreased efficiency. However, despite the obvious
advantages, in terms of efficiency in maintaining the turbine solidity below 0.5, it has
been found that low solidity turbines have poor self-starting characteristics compared to
the high solidity designs. It was shown experimentally and theoretically (Raghunathan
and Tan, 1983, Raghunathan et al., 1985) that besides the reduction in efficiency,
increases in σ also increase the pressure drop across the rotor. This important feature must
be taken into account since the pressure drop that the turbines can accommodate is a
major parameter to be considered in the design of a WEC. An alternative to the high
36
Chapter 2
solidity Wells turbine monoplane is a biplane turbine having two low solidity rotors
which can be used to provide acceptable levels of efficiency and self-starting ability.
The aspect ratio, AR, is defined as the ratio of the blade span, RT – RH, to the blade
chord length, c. According to Raghunathan (1995), the primary effect of reducing the AR
is to increase the efficiencies by postponement of stall; this is associated with the 'relief
effect' obtained on the blades due to relatively increased mass flow through the tip.
Moreover, the tip vortices become stronger with decreasing AR and this effect becomes
large with increasing tip clearance, tc.
The hub-to-tip ratio, h, is defined as the ratio of the hub radius, RH, to the tip radius,
RT. The effect of h on turbine efficiency based on experiments reported by Raghunathan
(1995) is shown in Figure 2.25. The results illustrate the combined effect of h and aspect
ratio, AR, since the tests were carried out by keeping solidity at a constant value of 0.4,
while varying AR. The importance of the hub-to-tip ratio in the design process is clearly
evident in Eq. (2.24), which demonstrates that for a given solidity and flow rate, Q, the
average axial airflow velocity, VA, depends on h2.
4Q
VA = . (2.24)
(
πD 1 − h 2
2
T )
Following Raghunathan (1995), the turbine aerodynamic efficiency can be affected
by h through a combination of the following factors:
• The airflow angle of incidence;
• The leakage losses at the tip;
• The relative interference effects at the hub.
For a turbine rotating at constant speed, the angle of incidence at the hub is always
larger than at the tip and becomes even larger with decreasing h. Further decreasing of h
may promote earlier stall onset and subsequently reduce the turbine efficiency. This is
one of the fundamental properties of the design variable h. Another essential feature of h
is related to the tip clearance, tc. Variation of h affects the tip leakage losses and
consequently the aerodynamic efficiency. Interference effects in vicinity of the hub are
also attributed to variation of h. Decreasing h reduces the interference effects produced
by the hub boundary layers and increases the efficiency. For design purposes the value of
h ≈ 0.6 has been recommended (Raghunathan, 1995).
37
Chapter 2
Figure 2.25 The combined effect of h and AR on efficiency for Wells turbines at σ = 0.4
(Raghunathan, 1995).
According to Raghunathan (1995) the Wells turbine performance is very sensitive to

variation of the tip clearance, tc. The onset of stall is advanced by decreasing the value of
tc, but at the same time the cycle efficiency is improved due to reduced leakage losses. On
the other hand increased tip clearance allows the Wells turbine to operate at higher angles
of incidence without stalling. However, increasing the value of tc, to more than 0.02 does
not demonstrate further improvement of the Wells turbine performance.
It is well known that turbomachines are quite sensitive to inlet conditions such as
turbulence level, Tu, and non-uniformity of the axial velocity profile. For instance, an
increase in turbulence levels can affect boundary layer development by advancing the
boundary layer transition, and delaying stall. However, it has been shown experimentally
(Raghunathan, 1995) that the Wells turbine is less sensitive to inlet turbulence as
compared to conventional axial flow air turbines. Experimental data were obtained by
testing a Wells turbine having the following design parameters: NACA0021 aerofoil
blades; turbine solidity σ=0.51; hub-to-tip ratio h=0.63; aspect ratio AR= 0.61 and Re =
2.8×105. The phenomenon of lower sensitivity of the Wells turbine is explained by its
high rotational speed (commonly 2000 - 3000RPM) which is significantly greater than
the axial flow velocity at inlet relative to which turbulence levels are measured.
Efficiency data in Figure 2.26 are plotted as a function of the flow coefficient, φ. It can be
seen that the increase of Tu more than by a half (from 3.2% to 6.7%), increased efficiency
38
Chapter 2
of the Wells turbine approximately by 3.6% only. In the present the range of Tu from 1%
to 10% as available options for the boundary conditions in ANSYS CFX-Pre will be
investigated to determine their effect on performance of the actual turbine models.
Figure 2.26 The effect of free stream turbulence on aerodynamics efficiencies

(Raghunathan, 1995).
Before presenting the effect of Mach number, M, on Wells turbine performance, it is

important to introduce Reynolds number, Re, and show its effect on the aerodynamic
forces acting on aerofoils commonly used for construction of the Wells turbine blades.
The Reynolds number, Re, represents the ratio of inertial forces to viscous forces. For the
purpose of the present study the following expression of Re will be used:
WR c
Re = , (2.25)
ν
where, WR is the relative airflow velocity on the blade, c is the chord length of the blade
and ν is the kinematic viscosity of air.
The effect of Re can be illustrated by presenting a typical variation of the tangential
force coefficient, Cθ, and axial force coefficient, Cx, as a function of the angle of
incidence, α, for symmetrical isolated aerofoils NACA00xx series published by
Raghunathan et al., (1985).
The aerodynamic force coefficients Cθ and Cx as functions of the incidence and
lift/drag coefficients (CL and CD) can be written as:
Cθ = C L sin α − CD cosα , (2.26)
39
Chapter 2
C x = C L cosα + C D sin α . (2.27)
Results shown in Figures 2.27 and 2.28 clearly demonstrate that maximum values of
aerodynamic force coefficients, Cθ, and, Cx, and the stall angle generally increase with
Re. It is also seen that the maximum values of Cθ and Cx as well as the onset of stall are
depended on the aerofoil thickness ratio, τ.
Figure 2.27 Variation of tangential force coefficient with the incidence (Raghunathan et
al., 1985).
From Figure 2.28 a linear relationship between Cx and pre-stall angles of incidence,
α, at both low and high Re is seen. This can be explained by the fact that in the pre-stall
region the axial force coefficient is largely dependent on the lift coefficient, CL, which is
a linear function of α. For a turbine operating at a constant speed, the angle of incidence
is also a function of flow rate and the pressure drop is proportional to Cx which suggests
the existence of a linear relationship between pressure drop and flow rate.
40
Chapter 2
Figure 2.28 Variation of axial force coefficient with the incidence (Raghunathan et al.,
1985).
As shown above increasing Re results in increases in both tangential, Cθ, and the
axial, Cx, force coefficients, respectively. It is desirable to keep the tip speed of the Wells
turbine at as high a value as possible since this provides more efficient aerodynamic
operation of the blades. However, the maximum tip speed must be limited in order to
maintain the blades rotation below the critical Mach number, Mcrt. If this condition is not
satisfied the flow over the blades becomes transonic leading to shock wave formation and
additional drag, which significantly exceeds the subsonic drag value. Moreover, the
shock wave and the boundary layer interaction on the blade surface results in a significant
increase in viscous drag. For this reason operation of the Wells turbine at tip speeds
where M > Mcrt is aerodynamically inefficient (Raghunathan and Beattie, 1996). In
addition, the unsteady pressure excitation associated with shock boundary layer
interaction in transonic flow results buffering and reduces the life of the blades. In order
to avoid transonic conditions: M ≤ Mcrt, where: M = VT/ca; VT is the blade tip velocity,
and ca is the speed of sound in atmospheric condition. According to Raghunathan (1995),
for a Wells turbine monoplane having a solidity of 0.3 and incidence range ± 8º, the tip
41
Chapter 2
Mach number, M, has to be less than 0.4. Therefore for design purposes a maximum
freestream Mach number M=0.4 can be adopted.
The critical Mach number Mcrt also plays an important role in design consideration of
the rotor geometry since as shown earlier, the tip diameter is determined by DT = 2VT/Ω,
where the value of VT is defined by the value of Mcrt.
Figure 2.29 The effects of h and σ on the self-starting ability of the Wells turbine. The
experimental data (closed symbols) indicate conditions for self-starting (Raghunathan,
1995).
In a wave energy converter one of the important design considerations is the

capability of the air turbine to self-start. The starting performance of the Wells turbine
was investigated by Raghunathan (1988), (1995). The findings demonstrate the effects of
hub-to-tip ratio, h, and solidity, σ, on the starting characteristics of the Wells turbine as
shown in Figure 2.29. It is clearly seen that low values of h and high values of σ must be
considered in order for a turbine to have good self-starting properties. The aerofoil
thickness ratio, τ also influences the starting properties of the Wells turbine. As
mentioned earlier, the thicker the blades, the better its self-starting ability.
The performance of the Wells turbine monoplane can be affected by a number of
design variables through their separate action as well as their combined effect. Taking
this information into account and using the principles given earlier some significant
conclusions can be formulated:
42
Chapter 2
• It is beneficial to extend the damping ratio BR = P*/φ into the stall region because
this tends to restrict the values of flow rate to be closer to the turbine’s efficient
range of performance; and
• It is aerodynamically inefficient to exceed the critical Mach number Mcrt ; and
• It is advisable to keep the angle of incidence both at the root and at the tip of the
blade below the stall onset; and
• It is desirable to minimise kinetic energy losses and recover the swirl energy
downstream of the rotor; and
• It is important to design the turbine with a good self-starting capability.
In the context of the current section, some of the given conclusions will be discussed
further along with demonstration of the Wells turbine development and feasible technical
solutions to improve its performance.
2.3.1.3 The biplane Wells turbine

As shown earlier, for the efficient performance of WECs, it is desirable that the
pressure drop across the turbine rotor is matched to the desired pneumatic damping
required to ensure that the OWC hydrodynamic capture efficiency is maximised. In the
case of a monoplane application it was shown that the pressure drop across the rotor of
the Wells turbine is approximately proportional to the square of the tip speed
(P*=∆P0/ρaΩ2DT2), which has to be less than the critical Mach number Mcrt if transonic
effects are to be avoided. Therefore, for WECs that are built to absorb wave amplitudes
(pressures) significantly larger than that dictated by the maximum tip speed of a Wells
turbine monoplane then a biplane design can be a solution.
The main idea of the biplane design schematically shown in Figure 2.30 lies in the
keeping of the overall turbine solidity at the level, which can accommodate the pressure
drops produced by the OWC by splitting the total solidity value between two turbine
rotors. As a result of the low solidity of each plane, such a design provides less
interference between blades and should lead to improvement in efficiency. Besides the
major design parameters described in previous sub-sections, the biplane configuration has
two additional characteristics, which determine relative positions of the planes, namely:
the gap to chord ratio, G/c, and the stagger angle, γ. Normally γ is equal to 0º, except in
the biplane designs with preset blade incidence. The parameter G on the biplane
schematic is the gap between rotors, c is the chord length, s is the spacing between blades
or cascade pitch, and s* is the shift between planes.
43
Chapter 2
A comparison of the efficiency of a monoplane turbine and a biplane (σ = 0.32 per

plane) turbine of equal overall solidity (0.64) is show in Figure 2.22 (Curran and Gato,
1997a) It can be seen that the damping ratio BR = P*/φ of the biplane is noticeably less
(by about 30%) than the monoplane. On the other hand, the biplane efficiency has
increased, although not by much. According to Curran and Gato (1997), who published
the results shown in Figure 2.22, the peak efficiency of the biplane was expected to be
higher as there was less blade work done due to the low solidity of the stages, and
consequently less swirl kinetic energy in the exhaust flow. Nevertheless, the main
achievement of the biplane configuration is seen in the stall region where the biplane
efficiency drops more gradually compared to the monoplane.
s*
c
s
Figure 2.30 Schematic of the Wells turbine biplane.
The results presented in Figure 2.22 for a biplane configuration based on the gap to
chord ratio G/c = 1.4. However, experiments carried out by Gato and Curran (1996)
showed that the variations of this design parameter leads to variations in turbine
efficiency, η, and non-dimensional pressure coefficient, P*. The effect of G/c on starting
and running characteristics of the biplane configuration was also studied by Kaneko et
al., (1991). The ratio G/c was varied from 0.3 to 1, and the best biplane performance was
achieved with G/c = 0.5.
In summary, the biplane configuration having the same overall solidity demonstrates
improved performance compared to the monoplane design, however, due to increased
losses produced by the downstream rotor the biplane efficiency increases very modestly.
The gap to tip ratio, G/c, is a very important design parameter affecting starting and
running properties of the biplane, and it must be properly optimised in order to provide
the best turbine performance.
44
Chapter 2
2.3.1.4 Fixed guide vane Wells turbine monoplane

Installation of fixed guide vanes before and after the rotor as shown in Figure 2.31 is
considered as one of the methods of recovering swirl energy, which may improve the
Wells turbine performance. Several publications (Curran, 2002, Inoue et al., 1985,
Curran and Gato, 1997a, Setoguchi et al., 2003b) have reported the investigation results
regarding the effect of the guide vanes. The basic idea in the use of guide vanes is to
optimally direct the airflow onto and off of the turbine blades, maintaining axial airflow
downstream the rotor and minimising air-swirl losses. The effect of guide vanes on
efficiency of the Wells turbine monoplane as well as on the non-dimensional pressure
coefficient, P*, based on the experimental results published by Curran and Gato (1997a)
are presented in Figures 2.32. Note: the data in this figure has been extracted from the
originally published curves by the present author (Curran and Gato, 1997a), which were
initially plotted separately to each other. This has been done intentionally to facilitate
direct comparison of the Wells turbine configurations with and without guide vanes. It is
seen that the rotor configuration with guide vanes maximises the turbine efficiency at
lower flow coefficients, φ, but decreases it at higher φ compared to the monoplane
configuration without guide vanes. Thus, despite the evident increase of the peak
efficiency, the performance of the monoplane Wells turbine with fixed guide vanes
rotating at constant speed can be optimised only for a limited range of flow rates
generated by OWC.
θ
Guide vanes
Rotor
Rotation
Guide vanes
Flow
Figure 2.31 Schematic of the Wells turbine rotor with guide vanes.
45
Chapter 2
On the other hand, a comparison of the non-dimensional pressure results clearly

indicate linear extension into the stall region of the damping ratio BR = P*/φ when the
Wells turbine rotor was equipped with the guide vanes. This turbine behaviour is likely to
be advantageous for the OWC and the overall system.
0.8 with GVs 1.0

with GVs
0.8
0.6
w/o GVs
w/o GVs
η 0.6
*
0.4 P
0.4
0.2
0.2
0.0 0.0
0.0 0.1 0.2 0.3 0.4
φ
Figure 2.32 Efficiency and non-dimensional pressure drop for a monoplane

configuration w/o and with guide vanes (replotted from Curran and Gato, 1997a).
It is useful to illustrate the effect of stagger angle of guide vanes on the Wells turbine
performance. This angle is indicated by the red circle in Figure 2.31. The influence of the
stagger angle of the guide vanes was studied by Inoue et al., (1985). Three different
angles: 7.8º, 11.8º and 15.8º were investigated and the best result in terms of turbine
efficiency obtained when the guide vanes were staggered at 11.8º.
In summary, fixed guide vanes are relatively simple equipment, which can improve
the performance of the Wells turbine, but prior to their installation modelling of the
stagger angle setting must be carried out in order to determine the optimum angle value
suited to the majority expected flow conditions.
2.3.1.5 The contra-rotating Wells turbine

Figure 2.32 shows that inclusion of the guide vanes increases the peak efficiency of
the monoplane by approximately 10% and noticeably extends the damping ratio BR into
the stall region, but adding the complexity of the turbine design. The biplane
configuration being less complex than the monoplane with guide vanes, also improves
46
Chapter 2
the Wells turbine performance, but at moderate scale due to the increased swirl energy
losses after downstream rotor.
The drawbacks associated with guide vanes and biplane design can be alleviated by
using the contra-rotating Wells turbine which consists of two rotors rotating in opposite
direction to each other (Raghunathan and Beattie, 1996). The basic idea of the contra-
rotating concept lies in elimination of the swirl produced by the upstream rotor by the
opposite rotation of the downstream rotor.
Typical variations of η and P* with φ based on data published by Curran and Gato
(1997a) for four Wells turbine configurations having the same overall solidity value of
0.64: a monoplane w/o guide vanes (MP), a monoplane with guide vanes (GV), a biplane
(BP), and contra-rotating design (CR) are presented in Figure 2.33. Note: data extracted
from originally published curves (Curran and Gato, 1997a), which were initially plotted
separately each other. It is done intentionally for direct comparison of different Wells
turbine designs. It can be seen that a contra-rotating Wells turbine is aerodynamically
more efficient than a biplane configuration. However, in the stall region when φ > 0.22
there is no noticeable difference between the efficiency curves for CR and BP. It is also
seen from this data that a contra-rotating configuration demonstrates the best
performance among the various Wells turbine designs.
0.9
CR
BP
0.6
MP
0.3
GV
0.0
0.0 0.1 0.2 0.3 0.4
φ
Figure 2.33 Efficiency for a monoplane w/o guide vanes (MP), a monoplane with guide
vanes (GV), a biplane (BP) and a contra-rotating configuration (CR).
47
Chapter 2
2.3.1.6 Air turbines with fixed non-zero stagger blade angles

In the previous part of this chapter the major design variations of air turbines with
blades fixed at a stagger angle of 0° to the plane of rotation have been presented. In this
section air turbines with fixed non-zero stagger blade angles are briefly summarised. One
driver for this approach is that data obtained in some OWC plants (Setoguchi et al.,
2003b), which show that the axial airflow velocity through the turbine in OWC system is
not equal in both directions. The results published by Setoguchi et al., (2002), (2003b)
derived from experiments and computer simulations also indicated that the axial flow
velocity during exhalation (i.e. flow from OWC chamber to atmosphere) is higher than
that during inhalation. Based on those findings it was proposed that the rotor blade pitch
be set asymmetrically at a positive stagger angle, γ, as shown in Figure 2.34, so as to
achieve a higher mean efficiency in a wave cycle. Figure 2.35 shows experimental results
reported by Setoguchi et al., (2002, 2003b) obtained under steady flow conditions, for a
Wells turbine without guide vanes with blades staggered at different angles (-4°, -2°, 0°,
+2° and +4°). It can be seen, that efficiency increases with the setting angle and reaches
the maximum when γ = 4°. The experiments also showed that the starting characteristics
of the turbine with positively staggered blades were better compared with blades fixed at
0°. The tests with the same set of stagger angles were carried out on the turbine with
guide vanes. The results demonstrated the similar trend of improved performance with
positive stagger angles. Testing of various stagger angles revealed that the value of +2°
was the optimum setting angle for both with and without guide vanes turbine
configurations.
Rotation γ
VA
Figure 2.34 Fixed non-zero blade angle setting, γ. Note the angle γ is measured from the
plane of blade rotation.
48
Chapter 2
Figure 2.35 Efficiency variations for a turbine without guide vanes with blades staggered
at different angles, γ (Setoguchi et al., 2002).
For completeness of the current topic, a schematic of a biplane Wells turbine with
blades staggered at angle γ is shown in Figure 2.36. The effect of stagger angles on
performance of the biplane turbine has been reported in articles by Kaneko et al., (1991)
and by Gato and Curran (1996).
VA
Rotation G
γ
VA
Figure 2.36 Schematic of the biplane Wells turbine with staggered blades.
49
Chapter 2
The optimum setting angle, γ, for the biplane design suggested by Kaneko et al.,
(1991) is in the range 2° ≤ γ ≤ 4°, which is close to the optimum stagger angle (γ = 2°) for
monoplane configuration proposed by Setoguchi et al, (2002, 2003b). Investigations
carried out by Gato and Curran (1996) showed that the biplane with staggered blades
increases the non-dimensional pressure coefficient, P*, compared to the design with
blades fixed at γ = 0°. At the same time the peak in the efficiency curve of the biplane
models with γ = 0° moved noticeably towards higher values of the flow coefficient, φ.
The experiments also revealed that for G/c ≤ 1, the adjacent upstream and downstream
pressure distributions will either suppress or enhance one another, depending on whether
they are aligned or staggered.
2.3.2 Air turbines with variable pitch angle blades

As shown in the previous section all modifications of the Wells turbines with fixed
pitch blades have their own unique operational advantages and working characteristics,
however, they perform well only over a limited range of airflows, or flow factor, φ.
Variable blade pitch, or variable geometry, turbines provide the opportunity for both the
range of airflow to be greatly increased at which high efficiency is achieved, and the
pneumatic damping of the turbine to be controlled. Both these effects can be potentially
achieved with a fast response rate to respond to both changes in air velocity during a
single wave cycle and also to longer term changes in the sea state. The variable geometry
approach allows control of the pressure and flow rate independently of each other and
hence it can facilitate the development of an optimal OWC control strategy.
2.3.2.1 The Wells turbine with self-pitch controlled blades

A design of the Wells turbine with self-pitch controlled blades (SPCB) is shown
schematically in Figure 2.37 has emerged as an attempt to overcome the drawbacks
associated with conventional Wells turbine configurations (Takao et al., 1997, Setoguchi
et al., 1997). The symmetrical aerofoil blades change stagger angle, γ, so as to obtain
higher torque in a reciprocating airflow. It is claimed that the SPCB design is simpler
geometrically and inexpensive to manufacture in comparison with the turbine proposed
by Salter (Salter, 1993) and Taylor and Caldwell (1998).
50
Chapter 2
VA
Pivot
Rotation
VA
Figure 2.37 Schematic of the Wells turbine with self-pitch controlled blades.
It has also been claimed that experimental and numerical data obtained during
investigations of the SPCB turbine (Kim et al., 2001, Takao et al., 1997, Setoguchi et al.,
1997) showed its superiority to the conventional Wells turbine both in the starting and
running characteristics. Analysis of the unsteady running characteristics carried out by
Setaguchi et al., (1997) revealed that the maximum mean efficiency of the SPCB turbine
was attained in case of γmax = 6°. However, despite the advantages claimed no published
data have been found in regards to the employment of an SPCB turbine design either in
the past or currently deployed OWC plants.
2.3.2.2 The Wells turbine with variable pitch angle blades

As mentioned earlier the use of a turbine with variable pitch angle blades (VPB) is
one of the potential methods of maximising the wave energy conversion through the
ability of control the pressure and the flow rate independently of each other. The concept
of the VPB design lies in adjustment of the blade stagger angle, γ, that goes
simultaneously with the changing airflow volume rate through the turbine as shown
schematically in Figure 2.38. The turbine blades are in a fixed position as long as the
incidence angle, α, is less than that required for stall (VA < WR sin αs). If the airflow
incidence exceeds this angle, as would be case with increased flow rate induced by the
larger amplitude waves (VA > WR sin αs), the blade stagger angle, γ, is adjusted to
maintain airflow at angle, α, which is less than the stall angle, αs.
51
Chapter 2
α < αs γ
WR
VA < WR sin αs VA > WR sin αs
α < αs WR
VA VA
Figure 2.38 A schematic diagram showing the concept of the VPB turbine.
Amongst the first VPB turbine configurations to be investigated was a modification

of the Wells turbine. Experimental and numerical results have been reported by several
researchers (Curran, 2002, Raghunathan, 1995, Salter, 1993, Gato et al., 1991, Tease,
2003). Analysis of published data clearly indicates that the VPB Wells turbine operates
over a much wider range of airflows at acceptable levels of efficiency compared to other
modifications of the Wells turbine based on fixed blade arrangement. One of the first full
scale VPB Wells turbines was developed to service the OWC system located on Pico
Island (Taylor and Caldwell, 1998). The rotor of the VPB turbine was designed to
accommodate 15 blades with a pitch angle range of - 40º to +40º (Figure 2.38).
Figure 2.39 Partially sectioned rotor of the Pico Island Wells turbine with variable pitch
blades (Taylor and Caldwell, 1998).
52
Chapter 2
The experimental results demonstrating the effect of the blade stagger angle, γ, on the
turbine efficiency obtained during the tests conducted by Wavegen (Tease, 2003) are
presented in Figure 2.40. It is evident that the increase of γ significantly extends the
efficient operational interval of the VPB Wells turbine modification. Some of the main
conclusions reported by Tease (2003) were:
• To maintain high energy conversion efficiency peak blade pitch rates of 32°/s are
required;
• Maximum blade pitch angles can be limited to ±30° without severe performance
drop off.
Figure 2.40 Effect of blade pitch angle on the turbine efficiency (Tease, 2003).
2.3.2.3 The Denniss-Auld turbine with variable rotor geometry

The Denniss-Auld turbine mentioned briefly in section 2.1.3 is another VPB air
turbine example developed by Oceanlinx Ltd. (previously known as Energetech
Australia) in collaboration with Doug Auld of the University of Sydney. There are a
number of features, which differentiate the design of the Denniss-Auld turbine from the
VPB Wells turbine. Firstly, the turbine blades are symmetrical about the mid-chord and
this property allows them to accept flow from either direction (Figure 2.41). Secondly,
the rotor blades can pitch relative to their mid-chord axis (Figure 2.42). As shown in
Figure 2.41, the stagger angle, γ, is defined as the angle between the plane of rotation and
the blade chord, c.
53
Chapter 2
Figure 2.41 Isometric view of the symmetrical about the mid-chord blade used in the
1/3rd scale Denniss-Auld turbine tested at the University of Sydney (Finnigan and Auld,
2003), where c is the chord length and b is the blade span.
Figure 2.42 Blade pitching sequence in oscillating flow (Finnigan and Auld, 2003).
The novel blade arrangement employed in the Denniss-Auld VPB turbine is claimed
to be suited to high-torque/low-speed operation in oscillating flows that vary from zero
flow to near transonic (Finnigan and Auld, 2003). It is interesting to note that the
Denniss-Auld turbine blades (Figure 2.42) are pitched in such a way that when the flow
switches direction the blades are required to rapidly flip across a large angle to accept the
opposite flow. According to Finnigan and Auld (2003), applying such a blade pitch
sequence provides the highest turbine efficiency over a wide spectrum of bi-directional
airflows. In addition, the turbine has extremely good self-starting characteristics.
Oceanlinx has published at least two different design options for actuation of the blade
movement cycle shown in Figure 2.42, i.e. the push-rod system shown in Figure 2.8 and a
bevel gear system described in PCT W O 2007/009163 Al (Denniss, 2007).
54
Chapter 2
Prior to building a full scale turbine the design concept of the Denniss-Auld VPB
turbine was tested on a 1/3rd scale prototype at the University of Sydney (Finnigan and
Auld, 2003). Experiments were carried out for blade stagger angles of 20°, 40°, 60° and
80° at a range of rotational speeds from 40rad/s [382RPM] to 230rad/s [2196RPM]. A
total of 59 tests were performed under different flow conditions through the range of the
Reynolds numbers from 1×105 to 6×105. The range of flow coefficients during the tests
was between 0.48 and 22.07. The maximum value of the flow coefficient was archived at
a stagger angle of 80°.
The experimental results shown by symbols in Figure 2.43 clearly demonstrate that
the operating flow range of the VPB turbine developed by Oceanlinx is considerably
wider than that of the conventional Wells turbine. The operating curve in Figure 2.42
intersects the efficiency peak at each tested pitch (stagger) angle indicating the optimal
value of γ for a variety of flow conditions. It should be noted that the operating efficiency
of the 1/3rd scale Denniss-Auld prototype was claimed to be greater than 35% even if the
flow coefficient, φ, exceeds a value of 5. Experiments also revealed that this VPB turbine
can generate positive torque even for a flow coefficient φ > 10 (Finnigan and Auld,
2003).
Figure 2.43 Operating efficiency curve based on data obtained from the tests of 1/3rd
scale Denniss-Auld prototype (Finnigan and Auld, 2003).
The findings presented by Finnigan and Auld (2003) and data obtained under the
present study using a blade element analysis and CFD modelling indicate that the design
of the Denniss-Auld turbine has the potential to increase operating efficiency over a large
range of air flows in comparison with other OWC air turbines. However, a number of
55
Chapter 2
design issues still remain to be resolved and optimised, which is the case for all types of
air turbines currently used or proposed for OWC systems. For example, the blade
activation mechanisms for both the Denniss-Auld and the VPB Wells turbines are
complex with many moving parts, which has significant implications for the longevity of
the plant in the harsh marine environment.
2.3.3 The impulse turbine

Although impulse turbines are outside the nominal scope of the present project a brief
summary of their characteristics for OWC applications is provided here. An impulse
turbine with fixed guide vanes introduced in section 2.1.3 was proposed by Kim et al.,
(1988) as an alternative to the Wells turbine in order to overcome drawbacks such as the
limited operational range. The impulse turbine can be a self-rectifying machine having
the fixed cup-shaped blades and a set of stator guide vanes upstream and downstream the
rotor, which act as nozzle and diffuser and vice versa under reversible airflows. The
performance of the impulse turbine depends on the same design parameters as the Wells
turbine such as the tip clearance, tc, hub to tip ratio, h, and Reynolds number, Re, etc.
Thakker and Dhanasekaran (2003) investigated the effect of tc on the performance of an
impulse turbine and found that the maximum efficiency is achieved when tc = 1%.
Efficiency variations with tc for the impulse turbine (Thakker and Dhanasekaran, 2003)
and results published by Raghunthan (1995) for the Wells turbine are shown in Figure
2.44. It is seen that both turbines are very sensitive to the value of tc, however, the
impulse turbine demonstrates somewhat less sensitivity than the Wells turbine.
The effect of hub-to-tip ratio, h, on the performance of impulse turbines have been
reported in several works (Thakker et al., 2002, Thakker et al., 2005a, Thakker and
Elhemry, 2007), and it would appear that the most efficient performance is attained by
the turbine configuration with h ~ 0.5.
56
Chapter 2
Figure 2.44 Effect of tip clearance, tc, on max efficiency of various turbines (Thakker
and Dhanasekaran, 2003).
Another important design parameter is the angle of the guide vanes, θ, indicated in
Figure 2.7 (see section 2.1.3) by the red circle. Maeda et al., (1999) studied five values of
the setting angle, θ, from 15º to 45º under irregular flow conditions and found that the
highest efficiency demonstrated by the impulse turbine with guide vanes, which were set
at 30º. It is interesting to compare performance of the Wells and impulse turbines under
comparable flow conditions. Variations of the mean efficiency, η , with the
dimensionless parameter, 1/(kω*), for both turbines are shown in Figure 2.45. The ratio
1/(kω*) characterises the irregular flow, where ω* is the dimensionless angular velocity,
and k is the dimensionless period of the oscillating air flow. It is evident that the impulse
turbine with GVs set at θ = 30º (solid curve) exhibited a higher mean efficiency over a
wider range of irregular flows compared to the Wells turbine (dashed curve).
In order to further improve the performance of impulse turbine, its basic design was
enhanced by employing self-pitch controlled guide vanes (SPCGVs) (Setoguchi et al.,
1991, Setoguchi et al., 1996). Here the SPCGVs rotate about pivots by means of the
aerodynamic moment induced by the reversible air flow (Figure 2.46).
57
Chapter 2
Figure 2.45 A comparison of the mean efficiencies simulated under irregular flow
conditions between the Wells turbine and the impulse turbine with guide vanes, θ =30º
(Maeda et al., 1999).
SPCGVs
Rotor Pivot
Rotation
SPCGVs
Flow
Figure 2.46 Schematic of an impulse turbine rotor with self-pitch controlled guide vanes
(SPCGVs).
The self-pitch controlled guide vanes variant of the impulse turbine (ISGV) shown in
Figure 2.46 was studied by Kim et al., (2001). Through the studies, other turbine types
were also investigated and their key performance characteristics under irregular flow
conditions were compared. Among them was a Wells turbine with guide vanes (WTGV),
58
Chapter 2
a turbine with self-pitch controlled blades (TSCB), a biplane Wells turbine with guide
vanes (BWGV), and an impulse turbine with fixed guide vanes (IFGV). Variations of the
conversion efficiency, ηm, with the dimensionless parameter 1/(kω*) published by authors
(Kim et al., 2001) for five turbines are presented in Figure 2.47.
The results in Figure 2.47 demonstrate that both impulse turbine types performed
better through the wide range of irregular flows compared to the Wells turbine types. The
maximum efficiency of about 47% was attained by ISGV turbine configuration and this
value was noticeably larger (about 15%) than the peak efficiency of the WTGV
configuration. Even for the simpler impulse turbine design (IFGV), the peak efficiency
was higher (about 6%) than attained by the WTGV turbine.
Figure 2.47 Conversion efficiencies of different turbine types (Kim et al., 2001).
The impulse turbine can work as either an axial or a radial machine depending on the
blade and rotor configuration as shown in Figure 2.48. The radial blade arrangement has
some advantages over the axial blades depositions, among them are reduction of the
oscillating axial thrust and low cost of manufacturing due to the simplicity of blade
geometry (Castro et al., 2007).
59
Chapter 2
a) b)
Figure 2.48 Impulse turbine configuration: a) radial and b) axial (Castro et al., 2007).
Takao et al., (2002) carried out numerical and experimental investigation on the
performance of radial impulse turbine under steady flow conditions using different
setting angles for inner and outer guide vanes. It was found that the optimum turbine
performance was attained when both inner and outer guide vanes were set at 25°. The
tests also revealed that turbine efficiency depends on the direction of the radial flow
through the turbine. The inward flow provides higher efficiency than the outward flow
direction.
Modelling and follow up analysis of the radial impulse turbine conducted by Castro et
al., (2007) gave the ground to claim that the radial design has a potential for significant
improvements in aerodynamic and mechanical characteristics compared to the axial
configuration. However, before the final dot is put on the current topic it should be noted
that no data have been found in open literature regarding the full-scale radial impulse
turbines employed for wave energy conversion.
2.3.3.1 Variable radius turbines

Amongst the most recent impulse turbines designed for OWC application is the so-
called Variable Radius Turbine (VRT), one example being the HydroAir turbine (Figure
2.49). This novel impulse turbine was developed by Dresser-Rand in conjunction with
Cranfield University, UK (Dresser-Rand, 2011). The VRT design comprises two sets of
static guide vanes located on either side and at a larger diameter than that of the rotor.
These vanes are connected by a converging-diverging duct to provide a route for the
airflow. Air enters the duct at a relatively low velocity and acquires a swirl motion as it
passes through the inlet guide vanes. The air then accelerates as it passes down the
60
Chapter 2
narrowing duct toward the turbine rotor. The air drives the rotor, and then decelerates as
it travels back through the expanding duct before passing over the outlet guide vanes. The
process is repeated (in reverse) for the next wave cycle.
Figure 2.49 HydroAir (Dresser-Rand, 2011)
It was claimed, that HydroAir offers the following benefits when compared to other
OWC turbines (Dresser-Rand, 2011):
• Above average efficiency
• One moving part - the rotor
• Lower rotational speeds than competing turbines
• Currently designed up to 500 kW
• Wide operating range
• Self-starting
• Reduced noise
2.3.4 Major OWC turbines tested in real sea conditions

The following table provides a list of some of the major OWC turbines that have been
tested under real sea conditions (Table 2.1).
Table 2.1 Major OWC turbines tested in real sea conditions

Rated
Type of Year of
Project power, Company Location
turbine deployment
kW
Wells
Limpet 75 Wavegen Isle of Islay, Scotland 1991
turbine
Bi-plane
Mighty
Wells 120 JAMSTEC Gokasho Bay, Japan 1998
Whale
turbine
61
Chapter 2
Wells
European Wave
turbine Island of Pico in the
pilot Pico 400 Energy 2005
with guide Azores, Portugal
plant Centre
vanes
Denniss- Port Kembla, 2005
Mk1 500 Oceanlinx
Auld Australia
Impulse Atlantic coast of 2006
OEBuoy No data OceanEnergy
turbine Ireland
2.3.5 Design methodologies: traditional and contemporary

The attainment of high efficiency by an air turbine built to service an OWC system
demands careful attention to details at the design stage. However, before the design stage
is actually commenced, an appropriate theory or analytical model of the turbine design
must be chosen. Design methods for OWC air turbines have evolved from methods
traditionally used for designing the gas turbines, axial flow fans and wind turbines which
were developed much earlier than their OWC counterparts. A literature review carried out
on the current topic revealed a number of methods, which have been used traditionally to
design the axial flow turbomachinery under steady state conditions. Amongst the most
frequently employed theories are: various empirical methods; free vortex radial
equilibrium methods; actuator disc and blade element momentum (BEM) methods. More
recently much more sophisticated design methods, including CFD, have been developed
to model complex three-dimensional (3D) effects occurring in rotors. An overview of the
more traditional design theories and modern design methods is presented below.
2.3.5.1 Free vortex radial equilibrium theory

The free vortex radial equilibrium theory of axial flow turbines was the dominant
design technique for many decades prior to the advent of computer-based analyses such
as CFD. The key assumptions of this method are that both the total pressure and the axial
velocity component remain constant along the blade length and that there is no radial
component to the flow. Following Wallis (1961), (1983) consider a small element of fluid
rotating relative the axis OZ in a pressure field at constant radius r and angular velocity ω
as shown Figure 2.50. When a balance between the centrifugal, Fc, and pressure, Fp,
forces acting on the element is maintained, a condition for radial equilibrium exists.
Forces Fc and Fp acting on the element having size, s', can be written as:
Fc = s′drρ
(ωr )2
, (2.27)
r
62
Chapter 2
Fp = dps′ , (2.28)
where, dp is the pressure difference between the two faces of the element. Equating the
forces Fc and Fp gives the universal requirement for radial equilibrium as follows:
2
dp
=ρ
(ωr )
. (2.29)
dr r
The total pressure of fluid element, P, in equilibrium can be expressed as:

1 1
P= p+ ρVA2 + ρ (ωr )2 , (2.30)
2 2
where, VA and ωr are the axial and tangential velocity components, respectively.
Y
s'
ωr
p + dp
p
dr
r
X
O
Figure 2.50 Rotating element of fluid (Wallis, 1961).
Differentiating Eq. (2.30) with respect to r gives:

2
dP dp 1 dVA2 1 d (ωr )
= + ρ + ρ . (2.31)
dr dr 2 dr 2 dr
Keeping P and VA constant with r, Eq. (2.31) reduces to:
2
dp 1 d (ωr )
=− ρ . (2.32)
dr 2 dr
Combining (2.29) and (2.32), in order to satisfy radial equilibrium requirements, it can be
shown that:
ωr2 = constant (2.33)
63
Chapter 2
i.e. ωr is inversely proportional to r.

However, this underlying assumption is valid only for the design the rotors having
untwisted blades with small aspect ratio (Turton, 1984). In relation to OWC turbines
radial equilibrium theory has been used for design and analysis by researches such as
Raghunathan et al., (1985), Raghunathan (1995).
2.3.5.2 Blade element momentum theory

Blade element momentum (BEM) theory is one of the oldest and most extensively
used methodologies for evaluating velocities and forces acting on wind turbine blades.
This theory being an extension of actuator disk approach combines two different theories:
blade element theory and momentum theory (Leishman, 2000). Applying BEM we
assume that blades are divided into small elements as shown in Figure 2.51 that act
separately from nearby elements and work aerodynamically as two-dimensional aerofoils
whose aerodynamic forces can be evaluated using the local flow conditions. Then these
elemental forces can be summed along the blade span to calculate the total forces acting
on each blade and to determine the moments exerted on the rotor. The second part of
BEM, the momentum theory, assumes that the loss of pressure or momentum across the
rotor is caused by the work done by the airflow passing through the rotor comprising the
blade elements. The momentum theory allows the evaluation of the induced velocities in
the axial and tangential directions from the momentum lost in the flow. These induced
velocities affect airflow through the rotor and thus also affect the forces evaluated by the
BEM theory. Combination of two theories in the BEM methodology sets up an iterative
process, which allows determining the aerodynamic forces and also the induced velocities
in vicinity of the rotor.
dr r
64
Chapter 2
Figure 2.51 Annular plane used in BEM theory.
2.4 CFD modelling of OWC turbines

There is no doubt that CFD has become a powerful tool for turbomachinery design
and analysis. Research and development of OWC turbines are areas where CFD is
applied quite extensively nowadays. The most significant benefit of employing CFD is
considerable savings in modelling time and in cost of the turbine design process. With the
help of CFD the well-known ‘trial-and-error method’ of physical building and testing of
prototypes prior to the final design can easily be avoided. Specific advantages of using
CFD for OWC turbines include: the ability to provide necessary flow data in the regions
that would be difficult or even impossible to test experimentally; the capability to
simulate real flow conditions over different aerofoil profiles within a short period of time
and compute the needed aerodynamic characteristics; the capacity to conduct tests on a
new air turbine design under real flow conditions and provide valuable data to its
optimisation; and also the ability to have visualisation of complex fluid flows in the area
of interest.
2.4.1 Historical overview

The application of CFD methods to OWC turbines dates back to the mid 1990s. The
first CFD models of the OWC turbines were relatively simple and often had rather coarse
meshes. Commonly the efficiency of the turbine model was calculated using the torque
and pressure drop data deduced from CFD simulations conducted under steady
incompressible flow conditions. Watterson and Raghunathan (1996) were among the first
researchers, who used the CFD method for analysis of the Wells turbine performance.
The authors built the computational domain with dimensions of 5 chord lengths upstream
and downstream of the blade row using a tetrahedral mesh, and restricted it to one blade-
to-blade passage, with rotational periodic boundaries along meridianal surfaces in the
same manner as shown in Figure 2.52. The CFD model had one NACA0015 blade and
overall rotor solidity σ = 0.5. For all CFD simulations, the rotational frame of reference
with speed Ω = 524 rad/s was set, which gave M=0.4, and Re = 8 × 105. CFD results
validated against experimental data (Gato and Falcao, 1988) showed that the CFD
method slightly underestimated the measured pressure coefficient, P*, of the Wells
turbine with σ = 0.59 for 0.05 ≤φ ≤ 0.15. The peak efficiency was also predicted to be
slightly lower. In addition to that the CFD method predicted the onset of stall at a smaller
65
Chapter 2
value of φ than observed in experiments. Watterson and Raghunathan (1998) continued

CFD modelling of the Wells turbine and investigated the effect of solidity on pressure
drop, torque and efficiency.
Inlet
Rotational periodic boundaries

Outlet
Figure 2.52 An example of the computational domain with rotational periodic

boundaries.
Davey and Tease (1999) further extended CFD analysis on a commercially developed
Wells turbine by applying the CFX 4 code. According to the authors, a good qualitative
agreement was demonstrated between CFD results and experimental measurements.
There are few works devoted to CFD modelling of the OWC turbines published
relatively recently. Thakker et al., (2001a) reported results of a CFD analysis of a Wells
turbine with CA9 blade profile. Also Kim et al., (2002) studied the effect of blade sweep
on the performance of a Wells turbine under steady flow condition with a full Navier –
Stokes solver for two blade types: NACA0020 and CA9. Setoguchi et al., (2003) using
FLUENT 5 investigated the effect of solidity, stagger angle and blade thickness on the
hysteretic characteristics and aerodynamic characteristics of a Wells turbine operating
under unsteady sinusoidal flow conditions.
Thakker et al., (2001b) carried out a 2D CFD analysis of an axial impulse turbine
with fixed guide vanes. The CFD model was meshed with Gambit and analysed using
FLUENT 5. Various turbulence models including the default k-ε, RNG k-ε and Reynolds
Stress Equation models were studied and results deduced from CFD simulations were
validated against experimental measurements. Since the CFD results were based on a 2D
analysis and tip gap losses were not taken into account, the turbine efficiency was
overestimated by all turbulence models tested in that study. The smallest difference
66
Chapter 2
between CFD and experimental efficiencies of order 6% was demonstrated by the k-ε
model. Thakker and Dhanasekaran (2003) further extended the axial impulse turbine
analysis by applying a 3D CFD model with tip clearance, which gave more realistic
predictions of the internal flow and turbine performance. The 3D CFD analysis employed
structured grids generated with Gambit 2 and the FLUENT 6 solver for problem analysis.
This study showed that the 3D CFD method overestimated the torque, CT, and input, CA,
coefficients of the turbine model approximately by 10% and 14%, respectively. At the
same time, the efficiency, η, was underestimated approximately by 3%.
Finnigan and Alcorn (2003) used CFD to calibrate the analytical model of the full
scale Denniss-Auld turbine with variable pitch angle blades. It is interesting to note that
the authors built the CFD model, which precisely represented the geometry of the turbine
hub, blades, and fore/aft diffuser sections.
Dhanasekaran and Govardhan (2005) conducted CFD analyses of the performance of
a Wells turbine by using FLUENT-6. In order to get the appropriate velocity profile at the
inlet as measured by experiments, the authors used a computational domain including
hub nose and all blades.
Thakker et al., (2005) used CFD to design and optimise the blade and guide vane
geometry of an axial flow impulse turbine and the authors also investigated the effect of
the hub-to-tip ratio, h, on the turbine performance on the basis of a 2D CFD analysis for
various φ. The main purpose of this research was to improve the performance of the
impulse turbine by modifying the shapes of blades and guide vanes as well as to
determine the optimum value of h with aid of CFD modelling. The influence of h on
impulse turbine performance was studied further with help of 3D CFD analysis (Thakker
and Elhemry, 2007). The analysis revealed that the turbine geometry with h = 0.5
demonstrated better performance compared to that with h values of 0.55 and 0.6.
Furthermore, 3D CFD analysis showed that the k-ε turbulence model can be used to
predict the performance of an impulse turbine for low rotational turbine speeds.
Among the most recent CFD studies of OWC turbines was the work by Torresi et al.,
(2007), who investigated influence of the hub nose geometry at the entrance and the tip
gap height on the performance of a small prototype of the high solidity Wells turbine.
CFD results reported in this work revealed only a marginal influence of hub-nose
geometry on the overall turbine performance, but confirmed its sensitivity from the width
of the tip gap. The effect of the tip gap height on the Wells turbine performance was
67
Chapter 2
studied further and CFD results presented in the paper by Torresi et al., (2008). This
paper reported a detailed 3D CFD analysis of the effect of TG variations on the turbine
performance as well as simulation results of the tip leakage flow inside the Wells turbine.
As in the previous research (Torresi et al., 2007), the CFD analysis was performed by
applying the commercial code FLUENT and solving the incompressible 3D Reynolds-
averaged Navier–Stokes (RANS) equations.
Castro et al., (2007) carried out CFD viscous flow analysis of the radial impulse
turbine by applying FLUENT. The research aim of the work was to improve the
understanding of the local flow behaviour and the performance prediction of the impulse
turbine having the radial configuration.
Gareev et al., (2009) validated the cascade interference factor predicted from the
inviscid flow method (Weinig, 1964) by applying 2D CFD simulations of the linear
cascades. The authors also conducted 3D simulations of a variable pitch Wells turbine
model using ANSYS CFX and compared CFD results with experimental data published
by Tease (2003).
This overview of the current topic using sources available in the public domain
clearly indicates that CFD methods have become important computational tools over
recent years helping the researchers to predict and optimise the performance of the OWC
turbines as well as to improve their designs.
2.4.2 Contemporary status

Initially CFD methods were mostly focused on the prediction of the turbine
performance characteristics, whereas now, with increased computational power, they are
used for more specific and more complex research and designing tasks. For instance,
Govardhan and Chauhan (2007) using FLUENT - 6.1 carried out steady incompressible
3D CFD analyses of two Wells turbine rotor configurations, namely, constant chord
(CONC) and variable chord (VARC). The VARC configuration with fixed guide vanes at
the inlet and outlet having circular arc profiles was also analysed. The VARC is a novel
rotor design aimed to overcome the drawbacks of the conventional Wells turbine
configuration with the CONC rotor. One of the major drawbacks of CONC geometry is
that the constant chord rotors exhibit flow separation in the hub region leading to strong
wakes and subsequent mixing of the flow at the downstream side of the rotor. In contrast
to CONC, the VARC rotor has more favourable angle of incidence angle along the length
of the blade. Besides the variable chord from hub to tip, the additional features of the
68
Chapter 2
VARC model were that their blades had a thicker profile at the hub (NACA0020) and a
thinner profile (NACA0010) at the tip. This was done to match the mechanical strength
requirements. For the VARC geometry with guide vanes, the computational domain was
divided into three volumes as shown in Figure 2.53.
Figure 2.53 Computational domain for VARC rotor with guide vanes (Govardhan and
Chauhan, 2007).
All CFD models were meshed by GAMBIT 2 applying the very fine mesh size to the
domain portion containing the blade and the coarse mesh to the domain parts located
upstream and downstream of the blade. Variation of power coefficient W* with φ deduced
from 3D CFD simulations for CONC and VARC rotors are presented in Figure 2.54. It is
seen that VARC rotor produces almost double the peak power output without showing
the stall onset compared to the CONC rotor.
69
Chapter 2
Figure 2.54 Variation of power coefficient, W*, with φ for CONC and VARC rotors
(Govardhan and Chauhan, 2007).
Figure 2.55 Effect of guide vanes on efficiency, η, of VARC rotor (Govardhan and
Chauhan, 2007).
The effect of guide vanes on efficiency, η, of the VARC rotor is illustrated in Figure
2.55. Efficiency simulations clearly indicate that the Wells turbine design with VARC
rotor is more efficient than the conventional turbine configuration with CONC rotor. It is
also seen that inclusion of guide vanes improves the efficiency of VARC rotor and it is
more noticeable at lower values of φ. Observing the results reported in regards to this
study, we see that applied CFD method has been able to provide the full 3D CFD analysis
of rather complex turbine geometries and generate essential data needed for their further
optimisation.
70
Chapter 2
Fixation of a generator
Plane 1
Plane 2
Figure 2.56 Two stage rotor setup with two rotating blade sections and one stationary
generator section with fixation of generator (Arlitt et al., 2007).
Arlitt et al., (2007) used CFD to model the aerodynamic behaviour of a biplane Wells
turbine with a stationary generator section between the rotating stages as shown in Figure
2.56. 3D CFD simulations gave insight into the flow interaction between the rotating
blades of the Plane 1 and the stationary generator section. It was visualised that due to
the large h the Plane 1 induced high twist near the hub causing flow separation at the
generator mountings, which needs to be avoided by optimisation of geometry.
Another description of CFD application for 3D analysis of complex turbine
configuration can be found in paper by Dorrell and Hsieh (2008). The authors used CFX-
5 to simulate steady state flow through the Wells biplane turbine model. CFD results
predicting the turbine performance were obtained at 1000rpm and 1500rpm through the
range of different steady-state inlet airflow velocities. It was found that turbine
performance characteristics deduced from 3D simulations of the CFD biplane model are
in line with data measured during the experiments of its small prototype. While the
prototype was a small demonstration unit the work illustrated the procedure and
calculations sequences that can be carried out in order to design and build a larger OWC
turbine with more practical application.
Contemporary computational methods are used not only to model the complex flows
and turbine configurations, but also different stresses caused by reversible and oscillating
flow through the turbine, which affect the their blades. For example, Tease et al., (2007)
performed a finite element (FE) analysis using ANSYS Workbench to find the stresses
acting on the Wells turbine blades under two load conditions. The first load condition
simulated the bending stresses as a result of a pressure drop (25kPa) across the blades.
The second load condition was a rotational speed (4000RPM). The results deduced from
71
Chapter 2
FE analysis helped to find the stress factor for the pressure loading and stress factor for
rotational speed. These data were further used for fatigue analysis and calculation of the
bearing life.
In summary, the modern computational methods are capable of modelling the
complex OWC turbine geometries and of simulating complex flows and other processes.
2.5 Concluding remarks

In the current chapter the classification of wave energy converters together with the
basic concepts of OWC technology have been presented. The variants of practical OWC
devices followed by consideration of appropriate air turbines designed to convert
pneumatic energy generated by WECs have been presented and discussed.
The importance of performance matching between the OWC and the air turbine has
been considered. The fundamental ideas, on which traditional and contemporary
methodologies for designing of OWC turbines are based, have also been given.
A parameter that significantly affects the hydrodynamic capture efficiency of an
OWC is the applied damping BA, which is provided by the air turbine. This parameter is a
function of important turbine design characteristics (BA = f(Ω, DT, AR, h, Np, σ).
It has also been shown that currently, the most frequently used variants of air turbines for
OWC application are self-rectifying axial air flow machines based on the invention of Dr
A.A. Wells. However, despite their design simplicity and robustness, the operational
range of these turbines is limited. On the other hand, impulse turbines and to a large
extent the Denniss-Auld turbine with variable pitch angle blades are able to operate over a
wider range of flow conditions as compared to Wells turbine types.
Although traditional methods for designing rotating flow machines are still widely
used, it has been shown that the trend of applying of CFD techniques for development of
modern turbomachinery including OWC turbines is becoming stronger. According to
Horlock and Deton (2005) nowadays the designers of rotating machines rely almost
completely on CFD to develop 3D blade sections. With the seemingly inevitable increase
in computing power, the future of CFD tools for design and analysis purposes looks very
promising. However, CFD will not be a complete substitute for experimental testing of
prototypes but rather will provide an aid to increase design efficiency.
Amongst the most recent and promising turbine configurations are OWC turbines
with variable rotor geometries. A particularly interesting design is the Denniss-Auld
turbine with variable pitch angle blades. Despite the fact that the Denniss-Auld design
72
Chapter 2
has been in existence for over a decade there remain a significant number of issues to be
resolved in terms of optimisation of its geometry and efficiency enhancement. The
present study was carried out in attempt to fill this gap in understanding and bring new
knowledge to the improvement of the design of the Denniss-Auld turbine.
73
Chapter 3
CHAPTER 3 - METHODOLOGIES FOR AERODYNAMIC

ANALYSIS OF OWC AXIAL FLOW TURBINES
This chapter provides a description of the blade element momentum (BEM)

methodology, which was used for performance evaluation of OWC axial flow turbines
analysed in the present study. Detailed description of the CFD analysis used for
prediction of key performance characteristics of OWC axial flow turbines modelled in the
present study is also presented and discussed.
3.1 Blade element momentum (BEM) analysis

This section gives a detailed description of the axial flow OWC turbine model based
on the blade element momentum theory introduced in chapter 2. The OWC turbine is
modelled as a ducted turbine. A single streamtube (Figure 3.1) and steady state
conditions are employed and compressibility effects, tip and frictional loses are not
included.
The BEM methodology for a ducted turbine provides a means by which the swirl
velocity downstream of the rotor can be determined. Thus, the forces on the individual
blades may be found from the lift and drag characteristics of the blades concerned, and
hence the pressure drops across the rotor as a whole determined.
∆P
VA RT
RH
V V
RH
Disc
Figure 3.1 Actuator disk and mass flow rate balance for a ducted turbine.
Since the cross-sectional area of the streamtube of the ducted OWC turbine is considered
constant (Figure 3.1), then the flow downstream of the rotor is restricted by the duct
walls, thus a mass flow rate balance can be expressed as:
( )
Ad V ρ a = A AV A ρ a = π RT2V ρ a = π RT2 − R H2 V A ρ a , (3.1)
74
Chapter 3
where Ad and AA are the cross sectional areas of the air duct and the annulus respectively,
V and VA are the axial air velocities across the duct and the turbine, respectively, ρa is the
air density, RT and RH are the tip radius and the hub radius accordingly.
From (3.1) and assuming no compression effects (ρa = constant), the axial velocity, VA,
across the turbine can be determined:
RT2
VA = V 2 . (3.2)
RT − R H2
The thrust (axial) force, FA, on the blades is equal to the rate of change in momentum as a
result of the pressure drop across the actuator disk, and on a per unit length basis is
defined by:
(
FA = ∆P × AD = ∆Pπ RT2 − RH2 , ) (3.3)
where AD is the actuator disc area and ∆P is the pressure drop across the turbine rotor.
The torque that results in the rotation of the blades induces an equal and opposite torque
on the air flow – i.e. swirl downstream of the rotor. From Newton’s Second Law of
Conservation of (angular) momentum, the torque is:
Torque = rate of change of angular momentum
= mass flow rate × change of tangential (swirl) velocity × radius
Following Sharpe (1990), the tangential velocity upstream of the disk is assumed
equal to zero, at the disk it is Ωra', and downstream of the actuator disk it is 2Ωra', where
a' is the tangential flow induction factor; which represents the fraction of tangential
velocity generated due to the rotation of the blades with the flow at radius, r. The torque
is evaluated at the average radius of the blades, which is given by:
 R −R 
r = RAV = RH + T H  .
 2 
Thus the torque acting on the blade is:
( )
Torque = ρ aV Aπ RT2 − RH2 × 2 a ′Ω R AV × R AV = 2πR AV
2
( )
ρ a VA a ′Ω RT2 − R H2 . (3.4)
Blade element theory gives rise to the blade element velocity vectors and forces on an
aerofoil, as defined in Figures 3.2 and 3.3.
75
Chapter 3
β
WR γ VT
α
VA
Figure 3.2 Blade element velocity vectors.
Where, VA is axial velocity at the blade, VT is tangential velocity at the blade
VT = ΩR AV (1 + a ') , WR is resultant relative velocity of the blade WR = V A2 + VT2 , γ is
stagger angle, β is angle with respect to the plane of rotation at which WR acts on the
blade, α is angle of attack α = (β − γ).
The forces acting on the blade can be determined as follows:
Torque
β
γ
α
Lift
Thrust
Drag
Figure 3.3 Forces acting on the blade.
Lift = CL × 0.5 ρ a WR2 × c Nb ,
Drag = C D × 0.5 ρ a WR2 × c Nb ,
76
Chapter 3
where, CL and CD are lift and drag coefficients accordingly; c is a chord length of the
blade; b is a blade span, (RT - RH); N is a number of blades. From Figure 3.3, the axial
force acting on the aerofoil blade can be expressed as:
FA = Lift cos β + Drag sin β = 0.5 ρ a WR2 Nc (C L cos β + C D sin β ) × (RT − RH ) . (3.5)
Similarly, the tangential component is:
FT = Lift sin β − Drag cos β = 0.5 ρ a WR2 Nc (C L sin β − CD cos β ) × (RT − RH ) . (3.6)
The torque on the blade can be written as:
Torque = 0.5 ρ a WR2 N c (C L sin β − C D cos β ) × (RT − RH ) × R AV . (3.7)
Solving (3.4) and (3.7) simultaneously gives an expression for the tangential
induction factor a':
WR2 N c (C L sin β − C D cos β )

a' = . (3.8)
4π R AV V A Ω (RT + RH )
If the lift and drag coefficients are not analytical functions of the angle of attack, then
the value of a' can be found by iteration of (3.8). First, the initial value for a' is defined
as zero a' = 0 and then velocity parameters (VT and WR) are obtained to solve equation
(3.8) and obtain a new value for a'. The new expression for a' is then inserted in the
expression for WR and the new values obtained are subsequently used to solve equation
(3.8) once more. This procedure is repeated until convergence occurs, that is a'i+1 - a'i =
itw, where itw represents the rate of convergence, i.e.: 0.05. As soon as a satisfactory
value for a' has been obtained, equations (3.3) and (3.5) can be simultaneously solved
producing the expression for pressure drop ∆P across the turbine rotor:
0.5 ρ a WR2 N c (C L cos β + C D sin β )

∆P = . (3.9)
π (RT + RH )
3.1.1 Dimensional characterisation of OWC turbines

Dimensional characterisation of OWC turbines is usually conducted by applying the
set of coefficients given below (Maeda et al., 1999, Takao et al., 2002, Setoguchi et al.,
2002, Finnigan and Auld, 2003, Thakker et al., 2005a, Torresi et al., 2007):
Pressure Coefficient:
∆P
P* = . (3.10)
ρ a Ω 2 R AV
2
77
Chapter 3
Input Coefficient:
∆PQ
CA = , (3.11)
( )
ρ a V A2 + [ΩR AV ]2 bcN
VA
2
where, Q is the volume flow rate in the duct and is given by Q = VA × π ( RT2 − RH2 ) . The
input coefficient, CA, provides a measure of the axial component of the lift and drag
forces acting on the blades, which gives rise to the rotor thrust.
Torque Coefficient:
Torque
CT = . (3.12)
( )
ρ a V A2 + [ΩR AV ]2 bcN
R AV
2
The torque coefficient, CT, provides a measure of the tangential component of the lift and
drag forces acting on the blades, which gives rise to the power output of the turbine.
Efficiency:
Power (out ) Torque × Ω
η= = . (3.13)
(
Power (in ) ∆P × π RT2 − RH2 × V A )
Flow Coefficient:
VA VA
φ= = . (3.14)
VT Ω R AV
3.2 CFD Analysis of OWC turbines

A brief introduction to the ANSYS CFX software package used for the purpose of the
present study followed by the short description of the CFD methodology adopted is
presented in this part of the chapter. The major elements on which CFD analysis is based
are also presented here, such as the governing equations, geometry modelling,
computational domain discretization, grid generation, boundary condition settings,
turbulence models and post-processing of CFD data.
3.2.1 ANSYS CFX

ANSYS CFX is a general purpose CFD program that has been applied to solve a wide
range of fluid flow problems for over 20 years. The ANSYS CFX package consists of
five software modules that exchange information required to perform a CFD analysis as
shown in Figure 3.4 (ANSYS, 2010). The creation of geometrical data for the CFD
simulations in this thesis was carried out using DesignModeller™, which was the part of
the ANSYS CFX software package.
78
Chapter 3
Figure 3.4 ANSYS CFX software modules that pass the information required to perform
a CFD analysis (ANSYS CFX Introduction, 2006).
ANSYS Workbench meshing mode provided access to various meshing tools in a

single module. The following mesh types were available: tetrahedral, hexahedral,
prismatic inflation layer, hexahedral inflation layer, hexahedral core, body fitted
Cartesian. ANSYS Meshing also had a physics preference setting to ensure that the right
mesh for CFD simulations was generated.
In order to setup the CFD analysis, the ANSYS CFX physics pre-processor was used.
Through the ANSYS pre-processor, flow physics, boundary conditions, initial values and
solver parameters were specified. The ANSYS CFX software had the capability to model
turbulent steady-state flows using many of the common Reynolds Averaged Navier-
Stokes (RANS) turbulence models such as k-ε, k-ω and Shear-Stress-Transport (SST).
The ANSYS CFX software came with specially tailored pre- and post-processing
environments that facilitated modelling of rotating turbomachinery as well as modelling
of complex machines with interaction between rotating and stationary components. The
present author was also able to use specialist turbomachinery modules, BladeModeler™
and TurboGrid™, geometry and mesh generation tools, respectively.
79
Chapter 3
3.2.2 CFD methodology adopted in the present study

The CFD methodology adopted in the present study comprised a number of pre- and
post-processing stages, which are summarised in Appendix A-4.
3.2.3 CFD governing equations

The governing equations used for simulation of the fluid flow are described in many
standard CFD texts (Abbott and Basco, 1989, Shaw, 1992, Versteeg and Malalasekera,
2007, Tu et al., 2008). A detailed description of CFD mathematics including all
governing equations is also given in the ANSYS CFX-Solver Theory Guide (2006).
Therefore only a brief outline of the key equations used for an isothermal viscous fluid
will be presented here. The equations are derived from the conservation of mass and
momentum (Jones and Clarke, 2003). The equation for the conservation of mass is
known as the continuity equation and can be expressed as:
∂ρ
+ ∇ ⋅ (ρU ) = 0 (3.15)
∂t
In (3.43), ∇ represents the vector operator for a Cartesian coordinate system, U is the
flow velocity, ρ is density, and t is time. If the fluid is assumed to be incompressible then
the density is constant and equation (3.15) can be simplified to:
∂ui
∇ ⋅U = =0 (3.16)
∂xi
where ui and xi represent the components of the velocity and position vectors, and the
summation convention for indices has been used. The equation representing the
conservation of linear momentum has the following form:
 ∂ui ∂ui  ∂p ∂τ ij
ρ  +uj =− + (3.17)
 ∂t ∂x j  ∂xi ∂x j
where τij is the viscous stress tensor and p is the dynamic pressure. For a Newtonian fluid
(which is also assumed throughout the present work) the viscous stress tensor can be
written in the form:
 ∂u ∂u 
τ ij = µ  i + i  (3.18)
 ∂x i ∂x j 
where µ is the dynamic viscosity of the fluid. Equation (3.46) can then be written as:
80
Chapter 3
 ∂ui ∂ui  ∂p ∂  ∂u j ∂ui 

ρ  +uj =− +µ + , (3.19)
 ∂t ∂x j  ∂xi ∂x j  ∂xi ∂x j 
which is known as the Navier-Stokes equation.
The momentum equation is given in the ANSYS CFX-Solver Theory Guide (2006) in
the following form:
∂ (ρU )
+ ∇ ⋅ (ρU ⊗ U ) = -∇p + ∇ ⋅τ + S M , (3.20)
∂t
where SM is the momentum source and the stress tensor, τ, is related to the strain rate by
 2 
τ = µ  ∇U + (∇U ) + δ∇ ⋅ U  ,
T
 3 
where T is the thermodynamic temperature and δ is the identity matrix or Kronecker
Delta function.
3.2.4 Domain discretization methods

Equations (3.15) and (3.19) form a set of partial differential equations for the
variables p and ui over a continuous domain representing a countless number of degrees
of freedom. To solve these equations computationally they first need to be transformed
into a discrete domain representing a finite number of degrees of freedom. For this
purpose three discretization methods are commonly used in CFD, namely: the finite
difference method, the finite element method and the finite volume method (Shaw, 1992).
ANSYS CFX-11 used the finite volume method which is similar to some extent to the
finite difference method and uses some features of the finite element method. A detailed
description of the finite volume method can be found in Versteeg and Malalasekera
(2007), Tu et al., (2008).
3.2.4.1 Discretization errors

Discretization errors lead to differences between the exact solution of the modelled
equations and a numerical solution with limited time and space resolution (e.g. Tu et al.,
2008). For a consistent discretization of the governing equations, the computed results are
expected to become closer to the exact solution of the modelled equations as the amount
of sub domains or grid cells is increased. However, the results are significantly affected
by the local density of the cells and their distribution.
Discretization errors are generated by localised sources and propagated (i.e.
amplified) throughout the solution domain. Localised sources of error can result from
81
Chapter 3
high-order terms that are excluded from the discretized the governing equations.
Reduction of the localised error sources can be achieved by increasing the order accuracy
of discrete approximations and/or by reducing the mesh spacing in regions of rapid
solution variation. For instance, applying the special mesh manipulation techniques (e.g.
point/lines control) available in ANSYS Mesh-mode, assists in avoiding poor geometrical
mesh quality in such regions and therefore helps to reduce solution errors.
3.2.5 Turbulence modelling

The equations governing fluid motion presented in section 3.2.3 describe the steady
laminar flow of a viscous, incompressible, Newtonian fluid without free-surface effects.
However, the majority of engineering tasks require the solution of complex flows of a
turbulent nature. Theoretically the Navier-Stokes equations can be applied to find
solutions for both laminar and turbulent flows without any additional information.
However, real turbulent flows at high Reynolds numbers comprise a considerable range
of turbulent length and time scales, which requires providing the size of the finite volume
element significantly smaller than it is practically achievable in modern CFD codes.
Moreover, the Direct Numerical Simulations (DNS) of real turbulent flows require
computational resources which are many orders of magnitude higher than available in the
foreseeable future (ANSYS CFX-Solver Theory Guide, 2006).
Turbulence models currently used in CFD are statistical models based on the
Reynolds Averaged Navier-Stokes (RANS) equations; they have been specially
developed to describe the turbulence flows without need to provide unthinkable fine
mesh and carry out direct numerical simulations.
3.2.5.1 Reynolds Averaged Navier-Stokes (RANS) equations

When looking at time intervals significantly larger than the time of turbulent
fluctuations, we can say that a turbulent flow exhibits average characteristics. The
original Navier-Stokes equations can be modified by the introduction of average and
fluctuating quantities, thus producing the Reynolds Averaged Navier-Stokes (RANS)
equations. The RANS equations represent the mean flow quantities only, while modelling
turbulence effects without a need for the resolution of the turbulent fluctuations.
Use of the RANS equations significantly reduces the computing power requirements
as compared to a DNS and is generally adopted by modern CFD codes for practical
engineering calculations. However, the averaging procedure introduces additional
82
Chapter 3
unknown terms containing products of the fluctuating quantities, which act like additional
stresses in the fluid. These terms, called “turbulent” or “Reynolds” stresses, are difficult
to determine directly and so become further unknowns. A comprehensive review of
turbulence modelling is given by Abbott and Basco, (1989). A detailed derivation of the
RANS equations is beyond the scope of the present study. Nevertheless, some essential
steps showing the process of their modification are presented here.
Firstly, the instantaneous velocity, U, can be expressed as an average component, U ,

and a time varying component, u as follows:
U =U +u , (3.21)
where U is given by:

t + ∆t
1
U=
∆t ∫ Udt ,
t
(3.22)
The parameter ∆t represents a time scale that is large relative to the turbulence
fluctuations, but is small relative to the time scale on which the equations are solved.
Secondly, substituting the averaged quantities into the original transport equations will
produce the RANS equations given below. Note that the bar is dropped for averaged
quantities, except for products of fluctuating quantities.
∂ρ
+ ∇ ⋅ (ρU ) = 0 , (3.15)
∂t
∂ (ρU )
+ ∇ ⋅ (ρU ⊗ U ) = ∇ ⋅ (τ − ρu ⊗ u ) + S M , (3.23)
∂t
where τ is the molecular stress tensor. It can be seen, that the continuity equation (3.15)
has not been altered, but the momentum and scalar transport equations contain turbulent
flux terms in addition to the molecular diffusive flux. Besides the steady state cases, the
RANS equations can be solved for transient flow simulations. They are sometimes called
URANS (Unsteady Reynolds Averaging Navier-Stokes equations). A comprehensive
description of all derivation steps of the RANS equations is given by Versteeg and
Malalasekera (2007).
The turbulence models available in ANSYS CFX-11 were divided into two main
classes: eddy viscosity models and Reynolds stress models. In the eddy viscosity models,
the turbulence was modelled as consisting of small eddies which are continuously
forming and dissipating, and in which the Reynolds stresses are assumed to be
proportional to mean velocity gradient. In contrast to that, the Reynolds stress models did
83
Chapter 3
not use the eddy viscosity hypothesis, but solved an equation for the transport of
Reynolds stresses in the fluid.
3.2.5.2 The k-ε turbulence model

Different types of turbulent flows require different applications of turbulence models.
In cases where knowledge of the expected turbulence flows is limited the starting point
for a CFD analysis is to offer the application of the standard k-ε turbulence model. Being
one of the eddy viscosity turbulence models, the standard k-ε model offers a relatively
simple level of closure since it has no dependence on the geometry or flow regime input.
For general purpose flow simulations, the k-ε model is said to offer a good compromise in
terms of accuracy and robustness (Versteeg and Malalasekera, 2007).
In the k-ε model, the turbulence velocity scale is determined from the local turbulent
kinetic energy, which in turn is determined from the solution of its transport equation.
The turbulent length scale is estimated from two properties of the turbulence field,
usually the turbulent kinetic energy k and its dissipation rate ε. The dissipation rate is also
determined from the solution of its transport equation. The turbulent kinetic energy k is
defined as the variance of the fluctuations in velocity and the turbulence eddy dissipation
ε is the rate at which the velocity fluctuations dissipate. The k-ε model in ANSYS CFX
introduces two new variables into the system of equations. The continuity equation (3.40)
is not changed
∂ρ
+ ∇ ⋅ (ρU ) = 0 , (3.15)
∂t
but the momentum equation is altered such that:
∂ (ρU )
+ ∇ ⋅ (ρU ⊗ U ) − ∇ ⋅ (µeff ∇U ) = −∇p′ + ∇ ⋅ (µeff ∇U ) + B
T
(3.24)
∂t
where B is the sum of body forces, µeff is the effective viscosity accounting for
turbulence, and p' is the modified pressure defined as:
2 2
p′ = p + ρk + µt ∇ ⋅ U ,
3 3
where µt is the turbulence viscosity.
Since the k-ε model is based on the eddy viscosity concept, then:
µeff = µ + µt . (3.25)
The k-ε model assumes that the µt is linked to the k and ε via the relation:
84
Chapter 3
k2
µt = C µ ρ , (3.26)
ε
where Cµ is the k-ε turbulence model constant.
The values of k and ε come directly from the differential transport equations for the
turbulence kinetic energy and turbulence dissipation rate. The full details are given in
ANSYS CFX-Solver Theory Guide (2006) and Versteeg and Malalasekera (2007).
Although the k-ε turbulence model has come to be widely used it is important to
recognise that it has some weaknesses. According to Tu et al. (2008), the standard k-ε
turbulence model performs poorly in flows having features such as flow separation, flow
reattachment, flow recovery, unconfined flows (e.g. free shear jet) and secondary flows in
complex geometrical configurations (e.g. flow around a popper valve). The main
limitation of the standard k-ε turbulence model is associated with its inability to predict
flows near the walls where the viscous laminar sub-layers and low Reynolds numbers are
dominant feature.
3.2.5.3 Alternative turbulence models

In order to overcome the limitations of the standard k-ε turbulence model, some
alternative turbulence models based on the on the eddy viscosity concept have been
developed over recent years. One of these alternatives is the RNG k-ε model, which is
based on renormalisation group (RNG) analysis of the RANS equations. The RNG
procedure systematically removes the small scales of motion from the governing
equations by expressing their effects in terms of large scale motions and modified
viscosity (Versteeg and Malalasekera, 2007). It is claimed that implementation of the
RNG procedure can improve the standard k-ε model.
Another alternative is the k-ω turbulence model developed by Wilcox (1988), which
uses the turbulence frequency ω = ε/k as the second variable. It solves two transport
equations, one for the turbulence kinetic energy, k, and one for the turbulent frequency,
ω. The stress tensor is calculated from the eddy-viscosity concept. The k-ω model
assumes that the turbulence viscosity, µt, is connected to k and ω via the relation:
k
µt = ρ . (3.27)
ω
k - equation:
85
Chapter 3
∂ (ρk )  µ  
+ ∇ ⋅ (ρUk ) = ∇ ⋅  µ + t ∇k  + Pk − β ′ρkω . (3.28)
∂t  σk  
ω – equation:
∂ (ρω )  µ   ω
+ ∇ ⋅ (ρUω ) = ∇ ⋅  µ + t ∇ω  + α Pk − βρω 2 , (3.29)
∂t  σω   k
where β′, α, β, σk and σω are the model constants and Pk is the production rate of
turbulence, which is computed as in the k-ε model. Full details are given in the ANSYS
CFX-Solver Theory Guide (2006).
The main advantage of the k-ω formulation over the standard k-ε model is the near
wall treatment for low Reynolds number computations. In addition, the k-ω model does
not involve the complex non-linear damping functions required for the k-ε model and is
therefore thought to be more accurate and more robust.
Despite evident improvements in the Wilcox model, it is still limited in its ability to
predict flow separation and its sensitivity to freestream conditions (Tu et al., 2008). In
order to mitigate the freestream sensitivity, Menter (1992a), (1994) proposed the baseline
(BSL) k-ω model, which combines the advantages of both standard k-ε and k-ω models.
The Menter model is a hybrid model using a form of the k-ω model in the near-wall
region and the standard k-ε model in the fully turbulent region far from the wall.
Although it is claimed that the BSL k-ω model is superior to the Wilcox model, it still
fails to properly predict the onset and amount of flow separation from smooth surfaces.
The main reason for this deficiency is that the BSL model does not account for the
transport of the turbulent shear stress. A detailed explanation of the drawbacks of the
BSL model is given in Menter (1992a, 1994).
The k-ω based Shear Stress Transport (SST) model was developed to overcome
limitations in the Wilcox and BSL models. The main distinguishing feature of the SST
model is that it includes the transport of the turbulent shear stress and gives highly
accurate predictions of the onset and the amount of flow separation under adverse
pressure gradients. The SST model is recommended for high accuracy boundary layer
simulations. To benefit from this model, a resolution of the boundary layer of more than
10 points is required (ANSYS CFS-Solver Theory Guide, 2006).
Since turbulence models based on the Reynolds stress concept have not been used for
the purposes of the present study, their explanation in the context of the current sub-
86
Chapter 3
section is omitted. The Large Eddy Simulation (LES) model and the Detached Eddy
Simulation (DES) model are also not explained here for the same reason.
In the present study, three turbulence models were tested and used for CFD
simulations, namely k-ε, k-ω and SST. Initial tests of these models for determination of
the lift and drag coefficients for isolated aerofoils revealed that the k-ε model had the best
prediction capability. For this reason the k-ε turbulence model was chosen for modelling
of aerofoils as well as for 3D CFD analysis of the OWC axial flow turbines discussed in
this thesis.
3.2.5.4 Near-wall treatments

In order to complete the turbulence modelling discussion it is important to discuss
problems associated with the modelling of flow near solid walls and the treatment of
near-wall turbulence. The major near-wall problems, which must be resolved in the
turbulence models, are viscous effects at the wall and quick variation of flow variables
which occurs within the boundary layer region. According to Cengel and Cimbala (2006),
turbulent flow in the near-wall region can be subdivided into four layers as illustrated in
Figure 3.5.
Figure 3.5 Representation of near-wall region (Cengel and Cimbala, 2006).
The closest layer to the wall is the viscous sub-layer, which is characterised by
domination of the molecular viscosity and laminar flow. Next to the viscous sub-layer is
the buffer layer, where the effects of molecular viscosity and the turbulence are of the
87
Chapter 3
same order of magnitude. Above the buffer layer is the overlap layer or transition layer,
in which the turbulence effects are much more significant, but still not dominant. The
turbulent layer is the furthest layer from the wall. In this layer the turbulence dominates
the mixing process. The assumption that a logarithmic profile is a reasonable
approximation to the velocity distribution in the near-wall region allows numerically
calculation of the shear stress as a function of the distance from the wall.
In the current version of ANSYS CFX-11 two options were available which could be
used for the treatment of near-wall regions: a) the k-ε turbulence model had a scalable
wall function and b) the k-ω and SST models were based on automatic wall treatment. In
addition to the turbulence models the near-wall regions could be resolved by applying
very thin inflation layers based on structured grid elements. Typical recommendations for
resolution of a boundary layer given in ANSYS CFX-Solver Theory Guide (2006) and
Tu et al., (2008) were to provide at least 10 and 15 nodes in the direction normal to the
wall. Based on these recommendations, the present author used the inflation layers in the
number between 10 and 15 inflation layers (depending on the size of the object modelled)
for all turbulence geometries tested in this study.
3.2.6 Modelling of CFD geometry

Modern CFD codes including ANSYS CFX offer possibilities to import existing
CAD models or create the solid model from scratch. In ANSYS CFX the function of the
geometry creation belongs to DesignModeller™ integrated into the “workbench”.
BladeModeler™ was another geometry creation tool incorporated in ANSYS Workbench
that was exclusively developed to design turbine blades and configuration of the turbine
rotors as shown in Figure 3.6.
L
lin
lout
Figure 3.6 Example of computational domain with rotational periodicity for a Wells
turbine model (γ = 24°).
88
Chapter 3
It must be remembered that the flow dynamics must be sufficiently developed across
the length, L, of the computational domain when the geometry for CFD simulations is
being created. In the example shown in Figure 3.6, it is vital to provide sufficient length
(lout) of the exit region in order to achieve the fully developed flow. Unfortunately, there
is no firm guidance in the literature as to what is the criteria of the optimum length of the
computational domain at the inlet (lin) and the outlet (lout) when the turbine of particular
configuration is being modelled. Based on the experience gained during this study, the
present author has adopted the following length criteria for CFD simulations of OWC
turbine models: lin= 3c to 5c and lout = 6c to 8c, where c is the chord length of the blade.
Since the scope of the study covers not only CFD analysis of the OWC turbines but also
different isolated aerofoils and their cascades, the domain length guidance in relation of
latter cases will be given when they are considered in the relevant chapters.
3.2.7 Mesh generation

Mesh generation is a vital stage in computing numerical solutions to the governing
partial differential equations of a CFD problem. The most important aspects of the mesh
generation process are the quality and quantity of the grid elements which subdivide
(discretizate) the computational domain into a number of smaller, non-overlapping sub-
domains.
The accuracy of a CFD solution is governed by the number of grid elements within
the computational domain. Prior to major flow simulations it is critical to define the
optimum number of the grid elements, which provides the acceptable level of numerical
accuracy and computational efficiency. This can be achieved by conducting grid
independence tests.
3.2.7.1 Types of grids

The grid elements typically used for the domain discretization are divided on:
structured and unstructured, depending on whether or not there exists a systematic pattern
of connectivity of the element points with their neighbours. As the name implies,
structured grid elements have some type of regular, coherent structure to the mesh layout
as presented in Figure 3.7. Unlike structured grid elements, the unstructured arrangement
of the elements is irregular and has no systematic pattern. The major advantage of the
unstructured elements is that they can be applied to complex geometries (see Figure 3.8),
where use of structured elements would have severe difficulty.
89
Chapter 3
Figure 3.7 Structured elements generated with aid of TuboGrid by the present author
for the Wells turbine with blades staggered at 32° relative to the plain of rotation.
Figure 3.8 An example of the complex geometry of the pneumatic system of an OWC
WEC including the turbine coupled to the nacelle meshed by the unstructured elements
(made by the present author).
90
Chapter 3
Figure 3.8 gives an example of the complex geometry such as the pneumatic system
of the OWC wave energy converter (WEC) meshed by the unstructured elements. The
unstructured elements built with aid of the CFX-Mesh, which is the general purpose grid
generator integrated into ANSYS Workbench. Besides the unstructured elements the
CFX-Mesh generator is also able to generate hybrid grids (structured and unstructured)
simultaneously (see Figure 3.9). Inflation layers shown in Figure 3.9 are an example of
structured grid elements, which are used for resolving the near-wall regions to capture the
viscous flow effects. In CFX-Mesh, the inflation layers presented in the near-wall vicinity
were created by using prisms of fine resolution normal to the wall but coarse parallel to
it. The relative thickness of the adjacent inflation layers is determined by a geometric
expansion factor, which is usually set between 1.0 and 1.5. Following various grid
refinement tests the present author adopted a value for the expansion factor of 1.1 for all
near-wall meshing cases.
Periodic pairs
````
Unstructured Inflation layers

elements
Periodic pairs
Figure 3.9 Example of a combination unstructured and structured grid elements around a
cascade of NACA0012 aerofoils.
The structured grid elements shown in Figure 3.7 are an example of domain
discretization for the Wells turbine model obtained by means of TurboGrid, which is
91
Chapter 3
another grid generator integrated into ANSYS Workbench. This tool was specifically
developed for meshing of rotating flow machinery by hexahedral meshes, which are used
in the ANSYS program to solve complex blade passage problems. However, despite the
quality of the mesh generated by TurboGrid, the present author has discovered that this
special grid generator had significant difficulty in creating meshes around blades
staggered at angles from 0º to 20º relative to the plane of rotation. In order to overcome
this limitation for these stagger angles, the author used the standard CFX-Mesh generator,
rather than TurboGrid to build unstructured elements in combination with inflation layers
as shown in Figure 3.9. This represented a significant extra workload but resulted in
successful modelling of the turbine over the full range of the blade stagger angles from 0º
to 20º. Nevertheless, TurboGrid was successfully used to generate structured grid
elements for other turbine models with blades staggered at 20º ≤ γ ≤ 90º tested in study.
3.2.7.2 Control features of the meshing process

When the most appropriate mesh type for the blade or turbine domain discretization
had been chosen, the next key issue was the determination of the mesh resolution
required in the each portion of the computational domain. In any CFD problem it is
crucial to generate a mesh, which resolves/captures gradients of concern (e.g. velocity,
pressure, and temperature gradients). The ANSYS CFX-Mesh generator had a number of
features, which assisted in the manipulation of the mesh in a particular region of the
computational domain. These features include body, face and edge spacing options and
other tools to refine the mesh.
Figure 3.10 shows an example of the type of computational domain covering the
entire pneumatic circuit of an OWC system as dealt with in Chapter 8. It can be seen that
the air flow makes a 90° turn on the way from the OWC chamber to the air duct and vice
versa. There is no doubt that such a sharp change of the flow direction must be
adequately resolved by the mesh in order to capture the relevant flow physics, such as
flow separation. Line Control, for example, was used as shown in Figure 3.10 to mesh
regions of concern with finer resolutions than the other less important portions of the
domain. This approach led to considerable savings in the use of computational resources.
92
Chapter 3
Radius of influence
Length scale
OWC chamber
Air duct
Line control
Inlet
``
Figure 3.10 The example of the mesh control of the region with anticipated high
gradients.
Some CFD models, especially those designed for turbomachinery, make effective use
of periodic pair boundary conditions, which force the flow leaving at one face to re-enter
at that face’s equivalent in a periodic mapping of the domain. The ANSYS CFX-Solver
was capable of making identical meshes on periodic boundaries. Typically two different
types of periodic boundary conditions are employed in CFD simulations, namely:
translation and rotational. An example of the translation type is illustrated in Figure 3.9,
which shows the meshing of a linear cascade of NACA0012 aerofoils. The blue solid
lines on Figure 3.9 represent the periodic pairs of boundaries. On the other hand, the
rotational type is applicable for rotating machinery. When the rotation option was
selected in ANSYS CFX, the axis of rotation based on two points in the domain had to be
specified.
3.2.7.3 Grid independence

As shown previously, the accuracy of the CFD solutions as well as the significance of
the discretization errors are strongly dependent on the size and quality of the mesh.
Therefore, it is very critical before the final CFD solution is obtained, to test whether it is
grid independent. In the present study, grid independence of all computational domains
93
Chapter 3
was tested by increasing the number of grid elements approximately by the factor of 1.2
until a solution was achieved where no significant changes in the results occur.
3.2.8 Selection of physics and fluid properties

Following meshing of the computational domain the next important step was to select
the flow physics and fluid properties. The flowchart (Figure 3.11) illustrates some of the
various flow physics option in ASYS CFX. In the present study, steady-state simulations
were carried out and following recommendations in the Best Practices Guide for
Turbomachinery (ANSYS CFX Reference Guide, 2006) the OWC turbines in the study
were analysed using the k-ε turbulence model. All flow problems in the present study
were modelled as incompressible.
Flow physics
Steady-state Transient
Viscous fluid
Turbulent flow
Compressible Incompressible
Internal (air duct) External (aerofoil)
Combined
(ducted turbine)
Figure 3.11 A flowchart presenting the flow physics according to the research objectives.
3.2.9 Modelling rotors and rotating machinery

One of the major goals of the present study was to carry out 3D CFD modelling of
OWC axial flow turbines. Modern CFD codes including ANSYS CFX-11 provide a
special option for 3D modelling of turbomachinery. Instead of modelling the full Wells
turbine with variable pitch blades, a simpler model comprising one flow passage with one
staggered blade can be simulated by applying rotational periodic boundaries (interfaces)
as shown in Figure 3.12. It can be seen that the overall three-dimensional computational
94
Chapter 3
domain (Figure 3.12) consists of three sections: a stationary domain with the inlet, a
rotational domain, which represents the rotor, and a stationary domain with outlet. The
stationary domains represent the motionless nacelle and air duct.
Rotational Domain
Inlet Stationary Domain

Rotational periodic interface
Outlet
Stationary Domain
General connection interface
Y Rotational periodic interface
X
Z
Figure 3.12 Setting up the rotational periodic and general connection interfaces for 3D
CFD modelling of the Wells turbine with blades staggered at 24°.
Besides the use of rotational periodic boundaries for the 3D CFD modelling of the
rotors and rotating machinery, it was important to correctly set the mechanism of the
frame change between rotational and stationary domains. There were three types of frame
change models available in ANSYS CFX-11: Frozen Rotor, Stage and Transient Rotor-
Stator. The Frozen Rotor model was the simplest solution, which did not perform any
circumferential averaging of the variables at the interface as the Stage model did.
Application of the Frozen Rotor assumed that the relative position of the components in
the rotational and stationary domains is fixed for the entire simulation. Additionally, this
model produced a steady-state solution for the axial flow turbine models having the
rotational and stationary frames of reference similar to shown in Figure 3.12. The major
disadvantage of this model was that the transient effect at the frame change interface
cannot be modelled.
The stage model was found to be best applied in the steady-state flow simulations
through multi-stage rotating machines. Initial considerations of the Transient Rotor-Stator
model revealed that this is the most powerful model available in ANSYS CFX-11.
Application of this model could provide more realistic interaction between rotational and
stationary the components through the sliding (frame change) interface. The transient
model was also able to predict the true transient interaction of the flow between a stator
and rotor passage. The main disadvantage of this method was that it required significantly
95
Chapter 3
larger computational resources as compared to the other models described above. Since
CFD analysis of OWC axial flow turbines in the present study did not involve either flow
simulations through the stator stages (with stationary cascades of blades) or rotating
multistage rotors, the Frozen Rotor model was adopted throughout.
3.2.10 Running the Solver

As soon as all necessary boundary conditions and interface models were set up, the
CFD solver was run, which included the steps outlined in Figure 3.13. The choice of the
solution control scheme was an important step to run the solver. There were three
solution control schemes available in Pre-processor of ANSYS CFX-11: First Order
Upwind Differencing Scheme (Upwind), High Resolution Scheme and Specify Blend
Scheme. Initial considerations of schemes revealed that the Upwind scheme was able to
provide the most robust performance with a “blend factor” of 0.0, but suffered from
numerical diffusion. Specifying the Blend Scheme allowed setting a blend factor between
0.0 and 1.0. In this scheme, the value of 0.0 was equivalent to using the 1st order
advection scheme and was the most robust option. On the other hand, a value of 1.0 used
the 2nd order differencing for the advection terms. With the High Resolution setting, the
blend factor values changed between 0.0 and 1.0 throughout the domain based on the
local solution field. In flow regions with low variable gradients, the blend factor was
maintained to be close to 1.0 for accuracy. On the other hand, in areas where the
gradients vary sharply, the blend factor was maintained to be closer to 0.0 for robustness.
After initial consideration of each scheme the High Resolution Scheme was chosen as a
final choice for all CFD simulations carried out in the present study.
96
Chapter 3
Initialisation
Solution control
CFD calculation
Check for convergence
Yes No
Stop Modify solution

parameters or mesh
Figure 3.13 An overview of the solution procedure.
In order to control the convergence of CFD solution a residual target of 1×10-4 was
used in all CFD simulations carried out in the present study. A comprehensive reference
to the issues of running CFD solver is Tu et al., (2008).
3.2.11 Verification and validation of CFD results

One of the most important stages of CFD modelling is the verification and validation
of the numerical results. Verification of CFD results can be defined as a process of
determining that a model implementation accurately represents the developer's
conceptual description of the model and the solution to the model (Versteeg and
Malalasekera, 2007)
On the other hand, the validation defined by Tu et al., (2008) is a process of assessing
of the CFD model uncertainty by using reliable experimental data and estimating the sign
and magnitude of the modelling error itself. In practice, assessment of the CFD results
can be carried out by establishing a range of physical conditions comparable with
experiments, and performing comparisons of the CFD and experimental data that span
the range of conditions.
In relation to the current research CFD simulations have been compared to
experimental data available in public domain as well as on the lab and full scale
measurements provided by Oceanlinx. In addition a preliminary mesh sensitivity study
97
Chapter 3
was carried out by the present author on a benchmark problem (orifice in a circular duct)
as described in Appendix A-1.
3.2.12 Post-processing of turbomachinery

The Turbo-mode post-processor of ANSYS CFX-11 had the capability to provide
direct calculation of important turbine parameters such as pressure drop across the rotor,
axial velocity at the rotor location, torque and rotational velocity of the blades. Besides
this, the ANSYS CFX Turbo post-processor based on the turbo coordinates also provided
turbomachinery-specific views of the data through meridional and blade-to-blade plots.
Figure 3.14 illustrates a typical 3D view of a Wells turbine model with a blade-to-blade
contour plot showing the pressure distribution through the computational domain at the
blade mid-span.
Inlet
Outlet
Figure 3.14 A typical 3D view of the Wells turbine model with a blade-to-blade contour
plot showing the pressure distribution through the computational domain at the blade
mid-span.
3.3 Design process overview for modern OWC turbines

From a designer’s standpoint it is important to have an explicit and straightforward
framework which can guide the designer through the complex process of modern OWC
turbine design, especially if a designer plans to use CFD methodology. Some very useful
recommendations regarding the design of OWC turbines have been reported by Cruz
(2008). However, there is no one page summary of all the essential steps that should be
98
Chapter 3
undertaken through the design process, particularly when CFD analysis is involved. In
this context the present author has tried to fill this gap by summarizing the key steps that
are essential for the practical design of modern axial-flow OWC turbines with aid of
contemporary CFD tools as follows:
o Determine the ocean wave energy resource characteristics of the chosen site.
o Develop the OWC geometry and deduce the overall system design based on
the wave resource available.
o Define design and off-design flow rates and pneumatic damping as the
starting point for turbine design.
o Determine the most promising turbine configurations through a range of BEM
analyses over full range of operating conditions and potential geometries (e.g.
blade profile, hub-to-tip ratio, etc).
o Conduct a number of detailed CFD analyses including:
• Choose initial turbine geometry (e.g. RT, RH, N, γ etc).
• Build turbine geometry using appropriate CFD tool
• Generate the most suitable mesh throughout the computational domain.
• Set necessary boundary conditions using appropriate CFD pre-processor.
• Run the CFD solver using steady state mode till solution is converged.
• Analyse CFD results using CFD post-processor.
• In case of unsatisfactory predictions, modify the initial settings or mesh
and re-run CFD calculations.
o Compare results based on BEM and CFD calculations.
o If possible compare numerical results with available experimental data from
pilot or full-scale plants.
o For optimisation of chosen turbine design, repeat detailed CFD analysis using
modified geometry. In principle this could be driven by a high-level
optimization algorithm.
The framework above was used by the present author as a basis for the research
conducted in this study. However, it should it be noted that the framework addresses only
the design of the turbine per se. Treatment of other OWC components (such as the
diffuser, nascelle, etc) should be considered in addition to the above.
99
Chapter 4
CHAPTER 4 - CFD ANALYSIS OF ISOLATED AEROFOILS

The Blade Element Momentum (BEM) method used for prediction of turbine
performance requires the input of aerodynamic aerofoil coefficients such as lift, CL, and
drag, CD. Thus, it is important that designers of OWC turbines have access to reliable
aerodynamic aerofoil data if they are to use the BEM method with confidence. Any errors
in the aerodynamic data can result in errors in the prediction of the turbine’s major
performance parameters. This, in turn, can lead to unrealistic economic estimates of the
success of OWC projects. The best case scenario is to find accurate experimental lift and
drag data; however, such data sets are often not available or even exist. To overcome this
problem, the designer has to use an appropriate tool to calculate the necessary
aerodynamic coefficients.
The advent of CFD together with high-speed computers has made the designer’s life
much easier providing a powerful means for complex aerodynamic calculations and
simulations. Modelling of aerodynamic coefficients needed for designing of
turbomachinery is one of the tasks that modern CFD codes are capable to fulfil
successfully. Here however, an issue of CFD results validation must be on the first place.
One of the aims of the present study was to investigate whether differences in the
aerodynamic data measured at the comparable flow conditions available in public domain
and to apply CFD modelling to generate the necessary lift/drag coefficients for the further
use in the BEM analysis of the actual turbines with variable rotor geometries. In order to
accomplish this task, first, experimental CL and CD coefficients of the aerofoil sections,
which are most frequently used to design the OWC turbines (e.g. Wells turbine) is
analysed, secondly a validation of some aerodynamic coefficients deduced from CFD
simulations against the experimental data is carried out. Final part of the chapter is
devoted to CFD modeling of the CL and CD used further in the present study.
4.1 Experimental CL and CD coefficients for various aerofoil profiles

Over the past few years a number of researchers have investigated available aerofoil
data sets and compared them against aerodynamic coefficients deduced from CFD
simulations. There are two main reasons for this: firstly, the experimental aerodynamic
coefficients published in different open sources demonstrate discrepancies and require
their adequate assessment; secondly, the designing of new turbomachinery requires the
100
Chapter 4
new lift/drag data, which can be generated with the aid of modern computational tools
including CFD, thus their accuracy must be validated prior the use.
Gretton and Bruce (2007), for example, used XFOIL, which is an interactive
computational program for the design and analysis of subsonic isolated aerofoils, to
calculate CL and CD for the NACA0012 airfoil profile. This aerofoil shape was taken as a
baseline for comparison of various experimental and numerical (XFOIL) aerodynamic
coefficients at a comparable range of Reynolds numbers. It was found that there were
considerable differences in the published CL and CD coefficients for the same aerofoil,
both in terms of experiments carried out and the numerical analyses. Variations of
published CL data at angle of attacks, α, from 0º to 32º for NACA0012 are shown in
Figure 4.1. It is clearly seen that the maximum attainable lift coefficients, CL, (before
stall) vary significantly. The same lack of agreement is also seen in the post-stall region.
Figure 4.1 Lift coefficients over the range of α from 0º to 32º for Re of around 2–3
million for NACA0012 airfoil (Gretton and Bruce, 2007).
101
Chapter 4
Interesting results (Fuglsang et al., 1998) relevant to the current topic are presented in
Figures 4.2 - 4.3. The CL/CD coefficients of an isolated NACA 63-215 aerofoil measured
in Riso National Laboratory, Denmark (RISO) compared to a XFOIL free transition
calculation and to a turbulent flow calculation obtained by applying a 2D Navier-Stokes
solver “EllipSys2D” at Re = 1.3×106. It can be seen that the CL curves measured and
calculated at low α are in good agreement (Figure 4.2). In contrast, at higher α, the
substantial difference between experimental and calculated data is noticeable, and this
difference increases as the α increases. It is evident from Figure 4.3 that the measured
and calculated CD data have the same general shape; but significant discrepancies
between all drag curves presented here are also evident. Though, the best prediction to
experimental data at α up to 12º is demonstrated by CD curve based on the XFOIL
calculation. It is also seen that the XFOIL data shows attached flow being maintained
until higher α and therefore the drag is underestimated in this interval. On the other hand
the turbulent flow calculations of EllipSys2D predict an increase in drag after separation
reasonably well compared to XFOIL. It can be concluded that the both calculations
methods (XFOIL, EllipSys2D) provides comparatively accurate prediction of the lift/drag
in the pre-stall region, but interpretation of calculated results in the post-stall region must
be conducted with care.
Figure 4.2 Measured CL curve compared to XFOIL free transition calculation and
EllipSys2D turbulent flow calculation at Re = 1.3×106 (Fuglsang et al., 1998).
102
Chapter 4
Figure 4.3 Measured CD curve compared to XFOIL free transition calculation and
EllipSys2D turbulent flow calculation at Re = 1.3×106 (Fuglsang et al., 1998).
As mentioned earlier, one of the aims of the present research was to investigate the
availability and validity of experimental CL and CD data published in open sources, which
could then be used for the validation of aerodynamic coefficients deduced from CFD
modelling. The investigation revealed a range of aerodynamic data for NACA00xx series
was available in the public domain (Loftin and Smith, 1949, Critzos et al., 1955, Sheldahl
and Klimas, 1981, Abbott and von Doenhoff, 1959). Figures 4.4 - 4.5 show the CL
coefficients measured for the NACA0012 aerofoil profile, which were grouped having
the comparable values of Re. Observation of CL coefficients clearly shows evident
difference between all experimental data in pre-stall region, and the difference increases
as the α increases. The same can be said regarding the lift data in the post-stall interval.
A comparison of experimental CL coefficients for NACA0012 aerofoil through the
angles of attack, α, from 0º to 90º taken from two different sources (Critzos et al., 1955,
Sheldahl and Klimas, 1981) is given in Figure 4.6. It can be seen that there are three
different regions on each CL curve. First there is a pre-stall region, roughly from 0º to 20º,
where all experimental data demonstrate relatively good agreement to each other. The
second region runs from 20º to approximately 30º, and it is a post-stall region.
Experimental CL coefficients here have a noticeable fluctuation. In the third interval from
30º to 90º, the CL curves exhibit similar behaviour and the average difference between
experimental results is ~ 6%.
103
Chapter 4
1.2 Loftin and Smith
Jacobs and Sherman
0.8
CL
Sheldahl and Klimas
0.4
0.0
0 10 20 30
αο
Figure 4.4 A comparison of CL for NACA 0012:(Jacobs and Sherman, 1937) at
Re=6.6×105; (Sheldahl and Klimas, 1981) at Re=7×105; (Loftin and Smith, 1949) at
Re=7×105.
1.5 Loftin and Smith

Critzos et al .
1.0
CL
Sheldahl and Klimas
0.5
0.0
0 4 8 12 16 20
αo
Figure 4.5 A comparison of CL for NACA 0012:(Loftin and Smith, 1949) at Re=2×106;
(Sheldahl and Klimas, 1981) at Re=2×106; (Critzos et al., 1955) at Re=1.8×106.
104
Chapter 4
1.5
Critzos et al .
1.0
CL
Sheldahl and Klimas
0.5
0.0
0 30 60 90
αο
Figure 4.6 A comparison of CL coefficients for NACA 0012:(Critzos et al., 1955) at
Re=1.8×106; (Sheldahl and Klimas, 1981) at Re=2×106.
2.4
Critzos et al .
1.8
CD
Sheldahl and Klimas
1.2
0.6
0.0
0 30 60 90 120 150 180
ο
α
Figure 4.7 A comparison of CD coefficients for NACA 0012:(Critzos et al., 1955) at
Re=1.8×106; (Sheldahl and Klimas, 1981) at Re=2×106.
105
Chapter 4
The CD coefficients for the NACA0012 aerofoil profile are compared in Figure 4.7
using data from Critzos et al. at Re = 1.8×106 and data from Sheldahl and Klimas at Re =
2×106. It is clearly seen that there are significant differences in the experimental CD
coefficients for the range of α between from 50º to 160º. Undoubtedly, such variations in
CL and CD can mislead the designers, and must be carefully verified prior their use.
The calculation tools (XFOIL, EllipSys2D) presented in this section demonstrated
their capability to provide acceptable prediction of lift/drag data in both pre- and post-
stall regions. This is especially important for the purpose of design and analysis of OWC
turbines, which must operate within a wide range of flow conditions including high
angles of incidence.
4.2 CFD modelling of aerodynamic coefficients for isolated aerofoils

Flow around an isolated aerofoil is one of the most fundamental problems in
aerodynamics. Measurement of forces acting on an aerofoil in a wind tunnel provides
valuable information that is used for determination of CL and CD. However, experimental
testing is a very time consuming and costly process. As demonstrated above, even
experimental measurements have a significant uncertainty associated with them. Thus,
CFD modelling of the flow field around an isolated aerofoil is a relatively convenient,
fast and inexpensive alternative providing that a benchmark CFD case can be validated
against reliable experimental data.
4.2.1 Calculation procedure for CL and CD coefficients

The lift and drag coefficients, CL and CD, are conventionally defined in terms of the
lift and drag forces, FL and FD, as
FL
CL = , (4.1)
0.5 ρV 2 A
FD
CD = , (4.2)
0.5 ρV 2 A
where ρ is the fluid density, V is the fluid velocity, and A is the planform area of the
aerofoil section.
4.2.2 Aerofoil NACA 64A010

The first aerofoil profile chosen to validate the CFD method in the present study was
the symmetrical NACA64A010 profile. Experimental CL and CD data, and profile
106
Chapter 4
coordinates were taken from the work by Selig et al., (1995). Since the range of α in
experiments was within the pre-stall region, the flow separation on the suction side of the
aerofoil during CFD modelling was not expected. Initial tests of domain dimensions
showed that in order to provide fully unrestricted flow around the aerofoil section, with
chord length c = 100mm, it was sufficient to locate the inlet and outlet four chord lengths
upstream and six chord lengths downstream, respectively. Taking this into account, the
final dimensions of the CFD domain used in study were 1000mm×1000mm×10mm. Prior
to carrying out CFD lift/drag modelling grid independence testing of the computational
domain was carried out. The test results are presented in Figures 4.8 - 4.9.
0.70
0.65
CL
0.60
0.55
0.50
1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9
6
n × 10
Figure 4.8 Dependence of CL coefficient from the number of elements, n.
107
Chapter 4
0.04
0.03
CD
0.02
0.01
0.00
1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9
6
n × 10
Figure 4.9 Dependence of CD coefficient from the number of elements, n.
Initial modelling of the aerofoil used α = 6.19º. A convergence criterion of 10-4 was
set for all CFD simulations carried out in the study. The unstructured elements in
combination with inflation layers were generated using standard CFX-Mesh generator
and their number was tested from 1.25×106 to 1.81×106. The tests showed that the
optimum number of elements, which did not substantially affect the numerical results,
was ∼1.7×106. Taking this into account the number of elements for all flow simulations
through the range of angles of attack as in experiments was kept around 1.7×106.
A great deal of work was carried out in the early stages of the development of the
CFD analysis on determining the optimum mesh topology. This work stood the present
author in good stead later in the project as a whole, as a clear understanding of the
implications of various mesh choices was gained. The best mesh for isolated aerofoils
was found to be one with fifteen inflation layers having an expansion ratio of 1.1, where
the thickness of the first inflation layer was kept between 0.08 - 0.1mm. The size of the
mesh in the near vicinity of each aerofoil was manipulated by applying a control line
between the leading and trailing edges. The point spacing on the control line was
generally set with the radius of influence equal to 30mm and a length scale of 2mm. In
addition, the grid density around the leading and trailing edges was increased by a point
control option with smaller settings. Both k-e and SST turbulence models were applied to
108
Chapter 4
obtain solutions for all flow simulations. The value of the inlet velocity for all runs with
the present scenario was set to 46.94m/s. This provided flow conditions comparable to
the experimental data range of the Reynolds number (Re = 304,100).
Prior to testing the aerofoil through the range of angles of attack as in experiments,
the effect of turbulence intensity at the inlet was determined. It was found that variation
of turbulence intensity from 1% to 10% had a negligible effect on the final output. Taking
this into account, the default value of 5% intensity was used in all CFD simulations. A
no-slip wall condition was imposed on the aerofoil surface. A static pressure of 0Pa was
specified at the outlet. The rest of a computational domain had a free slip wall boundary.
All CFD results presented in this chapter were obtained using a personal computer having
the following hardware specification: Intel (R) Core 2CPU 6700@2.66GHz and 2.6GHz,
2.99GB of RAM (since the same personal computer was used in many of the other CFD
simulations described in this thesis, the abbreviation PC is adopted for further
references). The “high resolution” advection scheme and “auto timescale” options were
used in all turbulence models run in the CFX solver.
0.9
0.6
0.3
0
CL
-8 -6 -4 -2 0 2 4 6 8
-0.3
-0.6
-0.9
αο
Figure 4.10 A comparison between experimental CL data for an isolated NACA64A010

(symbols: □-k-ε model, ●-SST model) obtained by the present author.
109
Chapter 4
A comparison between experimental CL and CD data and aerodynamic coefficients

deduced from the present CFD simulations are presented in Figures 4.10 to 4.11,
respectively. It can be seen that the both turbulence models (k-ε and SST) underestimated
the lift data measured at α: -6.25º, -0.87º, 0.36º, 3.28º and 4.64º (Figure 4.10). It is also
seen that the SST model predicted CL data slightly higher than were measured at α of -
4.24º, -2.53º and 7.74º. The average difference between experimental and predicted CL
coefficients was 3% and 7% for the k-ε and SST models, respectively. On the other hand,
both k-ε and SST turbulence models overestimated the drag coefficients, CD, at all angles
of attack tested in study except 7.74º, where the drag data were predicted lower than in
experiments (Figure 4.11). It was found that both models produced very close values of
CD between angles of -2.53º and 4.64º, but demonstrated noticeable difference for other
α. It is seen that the k-ε model predicted the drag data closer to the measured CD
coefficients than the SST model for all values of α tested in study. Estimation of predicted
CD data showed that they are different from the experimental CD coefficients
approximately by 10% and 12% for the k-ε and SST models respectively.
0.04
0.03
CD
0.02
0.01
0
-8 -6 -4 -2 0 2 4 6 8
α ο
Figure 4.11 A comparison between experimental CD data for an isolated NACA64A010

(symbols: □-k-ε model, ●-SST model) obtained by the present author.
110
Chapter 4
4.2.3 Aerofoil NACA 6409

In contrast to the NACA64A010 aerofoil, the NACA6409 profile is asymmetrical
about the chord. The CL and CD data of NACA6409 and profile coordinates were taken
from the same source as in previous example (Selig et al., 1995). The dimensions of a
computational domain, position of an aerofoil and its chord length were also the same as
in previous case. The number of grid elements for all flow simulations through the range
of α as in experiments (-2.89º, -1.38º, 0.21º, 1.66º, 4.95º, 6.94º, 8.08º, 10.1º and 11.3º)
was kept also around 1.6×106. The inflation layers and their parameters were set as
described for NACA64A010. The mesh refinement was attained by applying the same
control options and settings as given in previous example. However, contrast to an
aerofoil NACA64A010 for flow simulations over an aerofoil NACA6409 only a k-e
turbulence model was used to obtain solutions through the range of α. The boundary
conditions were the same as in previous case, except the value of inlet velocity. For all
CFD simulations the inlet velocity was set to 31.3m/s. This gave comparable results to
experimental data range of Reynolds number (Re = 203,100).
1.6
1.2
0.8
CL
0.4
0
-4 -2 0 2 4 6 8 10 12
α ο
Figure 4.12 A comparison between experimental CL data for an isolated NACA6409

(closed symbols) obtained by the present author.
Figure 4.12 illustrates experimental CL data (Selig et al., 1995) plotted as a solid line
with open symbols and aerodynamic coefficients deduced from CFD simulations by the
111
Chapter 4
present author (closed symbols). It is seen, that the k-ε turbulence model slightly
underestimated CL data for -2.89º ≤ α ≤ 6.94º. However, the lift coefficient at α = 8.08º
was predicted quite accurately, the difference between experimental and CFD results was
less then 1%. It is also seen that at higher values of α (10.1° and 11.3º) the lift data were
overestimated, but the onset of stall at α = 10.1º was captured quite well. Quantitative
estimation of the measured and modelled CL coefficients revealed that they differed
approximately by 2%.
0.05
0.04
0.03
CD
0.02
0.01
0
-4 -2 0 2 4 6 8 10 12
αο
Figure 4.13 A comparison between experimental CD data for an isolated NACA6409
aerofoil (solid line with open symbols) published by Selig et al., (1995) and CFD results
(closed symbols) obtained by the present author.
A comparison between experimental CD data (open symbols/solid line) and CD

coefficients (closed symbols) deduced from the CFD simulations carried out by the
present author is shown in Figure 4.13. It can be seen that the predicted CD data
overestimated the measured CD coefficients approximately by 15% for all values of
α except -2.89° where the CFD model produced a result lower on 5% than it was shown
in experiments. It is also noticeable that the difference between experimental and CFD
results increased with increase α.
112
Chapter 4
4.2.4 Aerofoil NACA 0012

Aerofoil NACA0012 is probably one of the most extensively investigated aerofoil
profiles in the history of aerodynamics. In the present study this aerofoil shape is also
extensively used in numerical and CFD models, both for basic validation purposes and
also as a profile in the Wells turbine. The aim of this sub-section is to make a comparison
between experimental CL and CD data and aerodynamic coefficients deduced from CFD
simulations.
CFD analysis of an isolated aerofoil NACA0012 was carried out using the same
dimensions of a computational domain, grid refinement techniques, boundary conditions
and settings as for two previous examples. The exception was only a value of inlet
velocity, which was set to 54.4m/s to give an appropriate Reynolds number (Re =
3.6×105).
A comparison between the experimental CL data for NACA0012 aerofoil taken from
two different sources (Sheldahl and Klimas, 1981) and (Jacobs and Sherman, 1937) and
CFD computed CL coefficients for angles of attack from 0º to 30º is shown in the Figure
4.14. The experimental lift data are presented as solid lines with closed symbols. CFD
results are plotted for three turbulence models: k-ε, k-ω and SST. Observation of these
data shows that in the pre-stall region from approximately 0º to 10º the experimental data
reported by Sheldahl and Klimas (1981) and the CFD results match closely. It is also
important to note that all turbulence models predicted the onset of stall as measured by
Sheldahl and Klimas (1981) at α = 10° quite accurately. However, the most accurate
prediction was achieved by the k-ε model, which produced a result on 0.1% higher than
in experiments. It is clearly seen from Figure 4.14 that experimental CL coefficients
reported by Jacobs and Sherman (1937) match to the lift data from Sheldahl and Klimas
and CFD data only in the interval of α between 0º and 2º. At higher α up to the onset of
stall, the lift data from Jacobs and Sherman have lower values compared to others. For
example, the peak value of CL from Jacobs and Sherman on 13% lower than it was
reported by Sheldahl and Klimas (1981) and predicted by the k-ε model. In the post-stall
region, when α varied from 10º to 30º, all predicted lift data demonstrated significant
difference from the CL reported by Sheldahl and Klimas (1981). In contrast to this, the k-
ε and SST models predicted the experimental CL data from Jacobs and Sherman relatively
accurately for 12º ≤ α ≤ 27º. The predicted CL coefficients in this post-stall region were
different approximately by 5% from the measured lift data (Jacobs and Sherman, 1937).
113
Chapter 4
1.2
0.9
C L 0.6
Sheldahl and Klimas
Jacobs and Sherman
0.3 k-epsilon
k-omega
SST
0.0
0 5 10 15 20 25 30 35
ο
α
Figure 4.14 A comparison of CL data for NACA0012 aerofoils at Re = 3.3×105 (Jacobs
and Sherman, 1937) and Re = 3.6×105 (Sheldahl and Klimas, 1981) and CFD analysis at
Re = 3.6×105 in the present study.
Figure 4.15 presents variations of CD coefficients based on experiments (Sheldahl and

Klimas, 1981) and deduced from CFD simulations at Re = 3.6×105 by applying the same
turbulence models as noted above. The experimental and CFD results are plotted through
the same range of α as in Figure 4.14. It can be seen that experimental and predicted CD
curves have the similar shape through the entire range of α investigated in study. It is also
seen that all CFD turbulence models overestimated the measured drag data approximately
by 70% for 3º ≤ α ≤ 12º. However, the drag coefficient, CD, at α =15º was predicted quite
accurately; the difference between experimental value and all CFD results was 2%. At
higher values of α up to 20º, the drag coefficient was best predicted by the k-ω turbulence
model, which showed only 1% difference from the measured CD. Other turbulence
models (k-ε and SST) overestimated the drag data in the same interval of α by 3% and
10% respectively. In contrast to this, the drag coefficient at α = 23º was better predicted
by the k-ε and SST models, which produced CFD results respectively on 2% lower and
4% higher than in experiments. At the α = 27º and α = 30º, the best prediction of CD was
114
Chapter 4
shown by the SST model, the difference between measured and CFD values was 0.2%
and 5% respectively. Other models (k-ε and k-ω) underestimated the measured value of
CD by 5% and 9% at α = 27º and by 8% and 13% at α = 30º respectively.
0.6
Sheldahl and Klimas
k-epsilon
0.4
k-omega
CD
SST
0.2
0.0
0 5 10 15 20 25 30
ο
α
Figure 4.15 A comparison of CD data for NACA0012 aerofoils at Re = 3.6×105 (Sheldahl
and Klimas, 1981) and CFD analysis at Re = 3.6×105 in the present study.
CFD analysis of all turbulence models (k-ε, k-ω and SST) tested in study revealed
their capability to predict the lift and drag data in the pre-stall region as well as the onset
of stall quite accurately. However, in the post-stall region, which is extremely difficult to
simulate with any great accuracy, the k-ε model predicted CL coefficients, which match
best of all to the experimental data from Jacobs and Sherman (1937) compared to other
turbulence models tested in study. Comparisons of the CFD results against the
experimental drag data reported by Sheldahl and Klimas (1981) in the post stall interval
also showed that the k-ε model provided the best prediction of the CD coefficients. Taking
into account the fact that the k-ε model requires less computational resources and less
time to obtain CFD results compared to others, this turbulence model was adopted to
generate the lift and drag coefficients for the practical aerofoils described in the next
section.
115
Chapter 4
4.3 CFD modelling of aerodynamic coefficients necessary for the

present study
In this section the CFD determination of the aerodynamic coefficients of two practical
aerofoils used for construction of the blades of a 1/3rd scale model and a full scale
prototype of the Denniss-Auld turbine design is described. The modelling was focussed
on generation of CL and CD data through the range of α from -90° to 90° for isolated
aerofoils. Furthermore, these CL and CD coefficients were later used for a BEM analysis
of both prototypes. An image of a full scale Denniss-Auld turbine is given in Figure 2.8
and that of the 1/3rd scale prototype turbine prepared for wind tunnel testing at the
University of Sydney is shown in Figure 4.16.
Figure 4.16 A 1/3rd scale prototype of the Denniss-Auld turbine design (Finnigan and
Auld, 2003).
4.3.1 Aerofoil for the1/3rd scale Denniss-Auld turbine

The first attempt at modelling and measuring of CL and CD coefficients through the
range of α from -90° to 90° for an isolated aerofoil used in the 1/3rd scale Denniss-Auld
turbine was reported by Finningan and Alcorn (2003). According to their report, the
coefficients over the productive (non-stalling) range of angles between -5° and 18° were
measured in a wind tunnel and beyond stall they were modelled by applying the Viterna-
Corrigan post-stall model (Lissaman, 1998) as shown in Figure 4.17.
116
Chapter 4
The profile of an aerofoil used in the 1/3rd scale Denniss-Auld turbine is formed by
combining two front halves of the NACA 65-418 profile. This combination provides
symmetry about the mid-chord shape of an aerofoil (see Figure 2.45), which in principle
allows acceptance of airflow from either direction.
Figure 4.17 CL and CD coefficients, as measured in wind tunnel tests (-5°< α <18°), and
as modelled (Finnigan and Alcorn, 2003).
Table 4.1 Mesh refinement settings for an aerofoil used in the 1/3rd scale Denniss-Auld
turbine
Length Radius of
Region Controlling
scale influence EF*
options
(mm) (mm)
Leading edge of an aerofoil Point 0.6 5 1.1
Trailing edge of an aerofoil Point 0.6 5 1.1
Suction/pressure sides of an aerofoil Line 2.5 45 1.1

Number of
Thickness of the first EF*
Inflation inflation
layer (mm)
layers
All regions listed above 15 0.1 1.1
*
EF - expansion factor
For CFD modelling of CL and CD coefficients, the computational domain with

dimensions (mm) of 1900×1000×15 was created. The boundary conditions were set as
117
Chapter 4
shown in Appendix A-5. Mesh refinement settings applied in vicinity of an aerofoil

section is listed in Table 4.1.The k-ε turbulence model with scalable wall function was
used for all flow simulations described in this section.
Prior to the CFD modelling of CL and CD coefficients through the range of α from -
90° to 90°, the grid independence tests were carried out for α = 90° and with an inlet
velocity of 92.4m/s that gave the value of Re=6×105. The number of elements was varied
by decreasing the body spacing parameter from the value of 30mm to 11mm. The results
of grid independence tests are presented in Table 4.2. It can be seen that the number of
2.09×106 (test 5) provided the obvious grid independence for the CFD solutions for both
CL and CD coefficients. Further increase of number of elements did not substantially
affect the CFD solutions. Taking this into account the average number of elements for all
CFD simulations was maintained as in the test 5.
Table 4.2 Grid independence tests of a aerofoil used in the 1/3rd scale Denniss-Auld
turbine
Test
Total number of elements CL CD
number
1 1.2×106 0.001 1.848
2 1.28×106 0.002 1.150
3 1.59×106 0.000 1.334
4 1.78×106 0.000 1.032
5 2.09×106 0.000 1.002
6 2.2×106 0.000 0.988
7 2.39×106 0.000 0.992
The values of inlet velocity for the CFD simulations were set to 15.4m/s, 46.9m/s and
92.4m/s so as to give Reynolds numbers of Re = 1×105, Re = 3×105 and Re = 6×105.
Two CL curves based on the CFD results (closed symbols/broken line) obtained in the
present study and lift data (open symbols/solid line) reported by Finnigan and Alcorn
(2003) are shown in Figure 4.18. Both sets of data were obtained at Re = 6×105. The red
circles on the graph indicate the interval of angles of attack (-5°< α <18°), where the CL
coefficients were measured in a wind tunnel (Finnigan and Alcorn, 2003). It was found
118
Chapter 4
that the k-ε turbulence model overestimated the measured CL data by approximately 12%
for α = -3°. On the other hand, the measured CL data were underestimated by
approximately 14% for 0° ≤α ≤ 5°. For other values of α (8°, 10°, 13°, 15° and 18°) lift
data were predicted quite accurately, the difference between measured and CFD results
was 1%. Analysis of both sets of lift data also revealed that the CL coefficients predicted
using the Viterna-Corrigan post-stall model and obtained in the present study are in close
agreement with an average difference of 5% for 60° ≤ α ≤ 90°. However, in contrast to
the lift data reported by Finnigan and Alcorn (2003) that show only one peak and the flat
stall on the lift curve, the CL curve based on CFD results demonstrated two peaks and the
deep stall between angles of 15° and 40°. Analysis of CL data showed significant
differences (on average ~70%) between predicted CL coefficients using Viterna-Corrigan
post-stall model and CFD method for -80° ≤ α ≤ -5°.
2
OLX lift data 18º
1.5
CFD lift data
0.5 -5º
CL
0
-90 -60 -30 0 30 60 90
-0.5
-1
-1.5
αο
Figure 4.18 The CL coefficients for an aerofoil used in the 1/3rd scale Denniss-Auld
turbine obtained in present study (closed symbols/ broken line) and lift data (open
symbols/solid line) reported by Finnigan and Alcorn (2003).
A comparison between CD coefficients obtained in the present study (closed symbols

/broken line) and drag data (open symbols/solid line) reported by Finnigan and Alcorn
(2003) is presented in Figure 4.19. Both sets of drag data also obtained at Re = 6×105.
The blue circles on the graph represent the interval of α between -5° and 18°. Analysis of
119
Chapter 4
CFD data in this interval showed that the k-ε model underestimated the measured CD
approximately by 44%. It is seen also that the CD data based on Viterna-Corrigan post-
stall model and CFD simulations demonstrated significant differences for -90º ≤α < -5º,
and the difference increased with decreasing α. A comparison of CD coefficients for α
equal to 25º, 30º and 70º revealed the good agreement between the drag data predicted by
both methods, the difference was 3%. The k-ε model predicted higher values of CD
(approximately by 13%) compared to the Viterna-Corrigan post-stall model for 30° < α <
70°. On the other hand, the CD coefficient deduced from CFD modelling at α = 80° was
on 7% lower than it was reported by Finnigan and Alcorn (2003).
2.1
1.4
CD
0.7
OLX drag data -5º 18º

CFD drag data
0
-90 -60 -30 0 30 60 90
αο
Figure 4.19 The CD coefficients for an aerofoil used in the 1/3rd scale Denniss-Auld
turbine obtained in present study (closed symbols/broken line) and lift data (open
symbols/solid line) reported by Finnigan and Alcorn (2003).
It is interesting to note that at α = 90° the CD coefficient obtained in the present CFD
analysis demonstrated a noticeable drop as compared to the drag data based on the
Viterna-Corrigan post-stall model. The considerable decrease of CD predicted by the CFD
model can be explained by strong flow recirculation on the suction side of the aerofoil
section positioned at α = 90° (Figure 4.20). It can be seen that downstream of the blade
there is a complex flow field with recirculation. The contour plot presenting the pressure
distribution in vicinity of an aerofoil section also clearly indicates symmetrical regions
120
Chapter 4
with negative pressure values that contribute to the appearing of vortices. The present
author’s opinion this is that due to back flow behind the aerofoil the drag force can be
decreased giving the smaller value of CD for α = 90°.
Tabulated CL and CD data deduced from CFD modelling for an isolated aerofoil used
in the 1/3rd scale Denniss-Auld turbine through the operational range of Reynolds
numbers are given in Appendix A-2.
Figure 4.20 Illustration of recirculation behind the aerofoil section positioned at α =

90°.The aerofoil profile used in the 1/3rd scale Denniss-Auld turbine.
4.3.2 Aerofoil for the full scale Denniss-Auld turbine

Before a description of the steps involved in the process of CFD modelling of CL and
CD coefficients for the aerofoil used in the full scale Denniss-Auld turbine, it is important
to mention that previous to the present study no measured or modelled aerodynamic
coefficients in relation to this aerofoil were available. The aerofoil for the full scale
Denniss-Auld turbine was also cambered and symmetric about the mid-chord (Figure
4.21).
121
Chapter 4
c
Figure 4.21 A circular arc profile blade as used in the full scale Denniss-Auld turbine.
CFD modelling of the CL and CD coefficients for this aerofoil type was conducted by
implementing the same approach as described above for a small aerofoil section.
However, due to the larger chord length of the aerofoil c=230mm, the domain size was
increased (with domain length up to 2900mm). Further increase of domain size did not
affect the numerical results; the changes were less than 0.1%. The boundary conditions
for all simulations were set as shown in Appendix A-5 and the k-ε model with scalable
wall functions was used to generate CL and CD coefficients. Settings used for the
refinement of the mesh in the vicinity of the circular arc profile aerofoil is given in Table
4.3
2000mm
3c
2900mm
Figure 4.22 The dimensions of a domain for modelling of flow over an isolated circular
arc profile aerofoil at α = -80°.
122
Chapter 4
Table 4.3 Mesh refinement settings for a circular arc profile aerofoil used in the full scale
Denniss-Auld turbine
Length Radius of
Region Controlling
scale influence EF*
options
(mm) (mm)
Leading edge of an aerofoil Point 1.25 5 1.1
Trailing edge of an aerofoil Point 1.25 5 1.1
Suction/pressure sides of an aerofoil Line 1.5 45 1.1

Number of
Thickness of the first EF*
Inflation inflation
layer (mm)
layers
All regions listed above 15 0.1 1.1
*
EF - expansion factor
Grid independence testing was carried out using the same method as described in the
previous section and results are presented in Table 4.4. It was found that 1.55×106 (test 3)
grid elements were sufficient to provide grid independent solutions for CL and CD. Thus,
the number of grid elements for all CFD simulations presented in this section was
maintained approximately as in test 3.
Table 4.4 Grid independence testing of a circular arc profile blade used in the full scale
Denniss-Auld turbine
Test
Total number of elements CL CD
number
1 8.58×105 0.486 0.021
2 1.25×106 0.484 0.020
3 1.55×106 0.488 0.021
4 1.8×106 0.488 0.021
5 1.86×106 0.489 0.021
6 2.12×106 0.488 0.021
7 2.55×106 0.488 0.021
123
Chapter 4
Lift data 2.0

Drag data
1.5
1.0
0.5
CL
0.0
-90 -60 -30 0 30 60 90
-0.5
-1.0
αο
Figure 4.23 The CL (open symbols/solid line) and CD (closed symbols/broken line) data
for a circular arc profile aerofoil obtained by CFD in present study at Re = 2.5×105.
The CL and CD data were generated through a range of α from -90° to 90° at
following Reynolds numbers: Re=2.5×105, Re=5×105, Re=7.5×105 and Re=1×106. These
flow conditions were achieved by setting the inlet velocity to the values 16.7m/s,
33.5m/s, 50.2m/s and 67.0m/s accordingly. The CL (symbols/solid line) and CD (closed
symbols/ broken line) data for a circular arc profile aerofoil obtained in the present study
at Re = 2.5×105 are illustrated in Figure 4.23. It can be seen that the lift data for the
circular arc aerofoil within the positive range of α exhibit the same trend as for the
aerofoil used in the small scale Denniss-Auld turbine. The CL curve between 0º and 90º
has two peaks and the first one is the higher than the second. Experimental CL data
obtained for a cambered steel plate in a similar range of Reynolds (Pandey et al., 1988) is
shown in Figure 4.24 for different ratios of f/c, where f is the concavity of aerofoil and c
is the chord length.
124
Chapter 4
Figure 4.24 Variation of CL coefficients with angle of attack for a cambered plate
(Pandey et al., 1988).
f/c=0.095 (CFD)
1.8 f/c=0.1(Pandey et al.,1988)
1.2
CL
0.6
0.0
0 30 60 90
-0.6
αο
Figure 4.25 A comparison between experimental (Pandey et al., 1988) and CFD (present
study) lift data for 0° ≤ α ≤ 90°.
A comparison between the experimental (Pandey et al., 1988) and the CFD (present
study) for 0° ≤ α ≤ 90° is shown in Figure 4.25. Experimental lift data presented in
125
Chapter 4
Figure 4.25 were measured for the cambered plate having f/c=0.1 at Re = 2.23×105. CFD
data were obtained with f/c = 0.095 and Re=2.5×105. It is evident that there is close
similarity in behaviour of the measured and CFD lift data. However, for 0° ≤ α ≤ 45°,
CFD lift data were noticeably lower (approximately on 29%) than data reported by
Pandey et al., (1988). On the other hand, both sets of CL coefficients match closely to
each other at higher values of α. The average difference between measured and CFD lift
data for 50° ≤ α ≤ 85° was less than 11%.
It should be noted that the CL coefficients obtained in the present study for negative
values of α indicate a rather complex behaviour (Figure 4.23). CFD analysis revealed that
the drag curve (closed symbols/broken line) shown in Figure 4.23 for -70° ≤ α ≤ 90° has
a shape similar to the drag curve plotted in Figure 4.19 for an aerofoil used in the small
scale turbine. However, for -90° ≤α < -70°, the values of CD noticeably decreased (Figure
4.23). A possible reason of such reaction was the flow recirculation behind the aerofoil as
shown in Figure 4.26.
Figure 4.26 Illustration of recirculation behind the circular arc profile aerofoil positioned
at α = -80°.
126
Chapter 4
Tabulated CL and CD data deduced from CFD modelling for an isolated circular arc
profile aerofoil through the operational range of Reynolds numbers are given in
Appendix A-3.

The successful use of CFD for the generation of lift and drag coefficients for isolated
aerofoils has been demonstrated in this chapter. CFD results have been obtained by
employing a number of turbulence models including the k-ε, k-ω and SST models. All
three demonstrated good match of CL data to measured coefficients in the pre-stall region
as well as good prediction of the onset of stall. The prediction of CD was less favourable
for all turbulence models tested in this study (as has been the case in the research of
previous authors). Nevertheless, the k-ε turbulence model provided a more accurate
prediction of both lift and drag data through the range of α tested. Moreover this
turbulence model is less time consuming and requires less computational power that the
others. These factors resulted in the choice of the k-ε model for subsequent modelling in
the present study.
One of the outcomes of the work described in this chapter was the generation of CL
and CD coefficients for two practical aerofoils used for construction of the blades of the
1/3rd and full scale Denniss-Auld turbines for a range of α from -90° to 90°. The use of
these aerodynamic coefficients in a BEM analysis of small and full scale Denniss-Auld
turbines is described in subsequent chapters.
It was found that the CL curve of a circular arc profile blade used in the full scale
Denniss-Auld turbine plotted as a function of positive angles of attack demonstrates the
same trend as experimental lift data for cambered plates reported by Pandey et al.,
(1988).
127
Chapter 5
CHAPTER 5 - TWO DIMENSIONAL LINEAR CASCADES AND

DETERMINATION OF INTERFERENCE FACTORS
As with the design of wind turbines, the most common starting point for the analysis
of rotors such as the Wells Turbine has been the blade element/actuator disc
methodology. Prior to the development of this methodology the analysis of axial flow
turbines was carried out by applying methods based on knowledge of the aerodynamics
of “cascades”, which involved a number of assumptions regarding the flow through the
rotor and a great deal of empirical data (see Dixon (1978), for example). A cascade is a
circular or linear series of blades (aerofoils), which can be characterised by mutual
aerodynamic interference between adjacent blades when the flow passes through them.
The major parameters responsible for the magnitude of mutual interference are the chord
length of the blades, c, and the spacing, s, between blades, which can also be expressed as
the cascade solidity, c/s.
One of the difficulties of the blade element model is that in a relatively high solidity
rotor, as is common in many OWC air turbines, the interference between the blades in the
circular cascade of blades is not known a priori. It is also true that fundamental
experimental testing of a cascade of aerofoils at high stagger angles, γ, (see Figure 5.1) is
virtually impossible in a wind tunnel (Raghunathan, 1988, 1995, (1996).
In the past mutual interference of adjacent blades in a cascade at various stagger
angles has been estimated by theoretical methods based on potential flow analysis. The
most widely known model is the inviscid analytical method for the flat or curved plates
arranged in a straight-line cascade as proposed by Weinig (1964). Another frequently
used potential flow method is the one where the surfaces of aerofoils in a cascade are
replaced by singularities (Raghunathan, 1995, 1996). A third approach reported is a semi-
empirical method based on a correlation between computed values of mean aerodynamic
force coefficients from a turbine test and 2D aerofoil data obtained in a wind tunnel
(Raghunathan, 1995, 1996).
Hawthorne and Horlock (1962) and others in the 1960’s pioneered the actuator
disk/blade element (BEM) method to determine turbine performance in a similar manner
to the analysis of un-ducted fans and turbines such as wind turbines. The blade
element/actuator disc methodology in a ducted turbine provides a means to determine the
swirl velocity downstream of the rotor and using lift/drag data for the blade profiles, the
128
Chapter 5
forces are calculated and hence the pressure drops across device as well as other
important turbine performance characteristics (CA, CT and η) are determined. Some of the
first researchers to apply the actuator disc methodology to the analysis of the Wells
turbine were Gato and Falcão (1984).
β
α
VT1 W1
VA
s
VT2 W2
c
VA
Figure 5.1 Nomenclature for a cascade of aerofoils with stagger angle, γ, on an axial flow
turbine rotor, with axial air velocity, VA, and tangential velocity relative to aerofoils, VT.
Today CFD may be used to obtain 2D cascade aerodynamic coefficients at any

stagger angle as well as to simulate the details of complex 3D airflow through an axial
turbine. However, the blade element/actuator disc analysis remains a key tool for the
conceptual stage of turbine design when the designer needs to obtain a quick and
reasonably accurate estimation of the turbine performance and to determine the possible
ways in which to optimise the configuration of a turbine with respect to parameters such
as radius, flowrates, rotational speed, etc.
Accurate prediction of turbine performance parameters using the blade
element/actuator disc methodology requires the input of reliable lift and drag data for
blades arranged in a cascade, as shown in Figure 5.1. Relatively little fundamental
research has been published on the flow in such rotors until recently. In the sections
129
Chapter 5
below some of the theoretical results of Weinig (1964) are presented. These have been
used by several researchers (Gato and Falcao, 1984, Gato and Falcao, 1988, Raghunathan
and Beattie, 1996) to modify isolated aerofoils lift coefficient so as to be applicable to the
analysis of cascades, as well as the results of Wong (1994), who used the method of
singularities to predict the cascade interference of the Wells turbines with different rotor
solidities are presented.
5.1 Analytical prediction of cascade interference

An infinite and linear cascade of aerofoils of a given arbitrary shape as depicted in
Figure 5.1 has practical interest for the present study. One of the reasons is that this type
of cascade is a simplified representation of realistic annular rotor of the Wells turbine
with staggered blades. The aerofoils in a cascade (Figure 5.1) interact with each other and
the CL and CD coefficients are no longer the same as for a single isolated aerofoil. The
ratio of the lift coefficient of a blade in a cascade, CL, relative to that of an isolated blade,
CL0 taken at the same angle of incidence, is known as the “interference factor”,
k0 = CL/CL0.
As mentioned above, testing of aerofoil cascades to determine k0 is extremely
difficult. Only a very few, small, data-sets applicable to turbines used in OWC
applications are available in the public domain (e.g., Raghunathan, 1988). Thus, most, if
not all, researchers analysing the Wells turbine have relied on the earlier work of Weinig
(1964) who determined an analytical prediction of the interference factor for a linear
cascade of infinitesimally thin flat blades using potential flow theory.
130
Chapter 5
Figure 5.2 Prediction of the interference factor, k0 = CL/CL0, from potential flow theory
(Weinig, 1964). γeff = 90° - γ, where the γeff is the stagger angle of the flat plates measured
from the axis of X.
Weinig’s results showed that the interference factor is independent of the angle of
incidence, α = β−γ , and is a function only of the stagger angle, γ, and ratio of s/c (where
s is the blade spacing, e.g. distance from leading edge to leading edge of adjacent blades,
and c is the blade chord length). The ratio of s/c is also referred as an inverse solidity. A
key graph from Weinig’s paper is reproduced in Figure 5.2.
In general for a cascade with arbitrary solidity and stagger angle Weinig’s inviscid
flow analysis provides the following estimate of lift coefficient for an isolated aerofoil
prior to stall modified by the interference factor, k0:
CL = 2πk0sin(αm), (5.1)
where αm is the angle of attack (based on the mean of the velocities upstream and
downstream of the cascade) and where k0 (shown in Figure 5.2) was found through the
solution of a set of algebraic equations (Weinig, 1964). In the case where the stagger
angle γ = 0°, which is applicable to the case of the Wells turbine, Weinig gives the
interference factor as:
2s  πc 
k0 = tan  , (5.2)
πc  2s 
131
Chapter 5
where s is the spacing between blades and c is the chord length. Equation (5.2) has been
extensively used by researchers as the correction factor to modify the lift coefficient of
fixed-pitch Wells turbine blades (see, Figure 5.4).
Figure 5.3 illustrates two lift curves plotted as functions of angle of attack. The solid
line is based on the isolated lift data for NACA0012 obtained at Re = 7×105 (Sheldahl
and Klimas, 1981) and the broken line represents the cascade lift data corrected by
applying (5.2). In order to determine k0, and plot a cascade lift curve, the following
cascade dimensions namely: c = 0.1m and s = 0.169m were used in Eq. (5.2). It is clearly
seen that the lift curve (broken line) based on the lift data modified by k0 exhibits the
higher values of CL than the lift curve (solid line) based on the isolated lift data. However,
it should be noted that Weinig’s inviscid method makes it possible to estimate cascade lift
coefficients, but not the values for drag. For BEM analysis, the cascade drag data are
assumed to be equal to the values of the drag coefficients obtained for an isolated aerofoil
(Raghunathan, 1995, 1996).
1.8
Cascade lift data NACA0012
CLc
1.2
CL
0.6 CLi
Isolated lift data NACA0012

(Sheldahl and Klimas)
0
0 5 10 15 20 25 30
α ο
Figure 5.3 Representation of isolated lift data CL0 (solid line) based on Sheldahl and
Klimas (1981) at Re = 7×105 and predicted cascade lift data CL corrected by k0 from (5.2)
and assuming s/c = 1.69.
132
Chapter 5
Another analytical approach to the evaluation of the mutual interference between

aerofoils in a linear cascade is the method of singularities. This method, as the previous
one can predict only cascade lift values. The drag can be determined by a boundary layer
analysis using the pressure distribution obtained by the potential flow method. According
to Raghunathan (1995, 1996), the aerofoils in a cascade at a given angle of incidence
have an equivalent incidence, αeq, that produces approximately same pressure
distributions on the upper aerofoil surface as that of an isolated aerofoil at an incidence,
α. It is suggested that both the tangential force, Cθ, and the axial force, Cx, coefficients
for an aerofoil in a cascade can be evaluated by a simple proportional formula:
C C C 
θ = x =  x  pred , (5.3)
x0  x0 
C C C
θ0
where, Cθ0 and Cx0 are the force coefficients on an isolated aerofoil (e.g. measured in a
C 
wind tunnel) at an equivalent angle of incidence, αeq;  x  pred is the predicted ratio of
 C x0 
the normal aerodynamic force on an aerofoil in a cascade, Cx to the normal aerodynamic

force on an isolated aerofoil, Cx0, at the angle of incidence as calculated by the potential
flow method (Raghunathan, 1995, 1996).
One of the researchers who used the method of singularities known also as the
Martensen method for analytical prediction of cascade interference for aerofoils arranged
in a linear cascade as depicted in Figure 5.4 was Wong (1994). Figure 5.5 demonstrates
Wong’s results for normalised CL on a NACA0012 aerofoil plotted through the range of
solidities (or ratio of c/s). It can be seen that the normalised CL is the ratio of CL/CL0
which is the same as a cascade interference factor, k0, introduced by Weinig.
133
Chapter 5
αm
W1 α1 Wm
VT1
VA s
W2
α2
VT2
c
VA
Figure 5.4 Nomenclature for a cascade of aerofoils with stagger angle, γ = 0°, on an axial
flow turbine rotor with chord length, c, and pitch between blades, s; αm = (α1 + α2)/2;
Wm= (W1+W2)/2.
Analysis of results from the Martensen method in Figure 5.5 clearly showed that
increases in solidity, c/s, result in an increase in normalised CL. Moreover, a non-linear
relationship was observed between normalised CL and solidity.
The relationship between normalised axial force coefficient, Cx, and solidity reported
by Wong (1994) is presented in Figure 5.6. The predicted results of Cx was also
compared against the Weinig prediction and experimental data (Raghunathan et al.,
1990). It can be seen that the Martensen prediction over-estimated the value of
normalised axial force coefficient, Cx compared to both Weinig and experimental data.
134
Chapter 5
Figure 5.5 Normalised values of CL for a NACA0012 aerofoil at different solidities

(Wong, 1994).
Figure 5.6 Normalised values of Cx for a NACA0012 aerofoil at different solidities

(Wong, 1994).
135
Chapter 5
Returning back to Weinig’s results, it is important to mention that despite the fact that
(5.2) has been widely used by researchers as a correction factor to be applied to Wells
turbines (Gato and Falcao, 1988, Raghunathan and Beattie, 1996) there has been very
little, if any, validation of this relationship between k0 and solidity for practical aerofoil
cascades. Moreover, the more general relationship between interference factor and
solidity for cascades of arbitrary stagger angle (Figure 5.2) has received virtually no
attention in the public domain literature to date.
In the next section, comparisons of Weinig’s theory and calculations reported by
Wong (1994) with CFD results obtained in the present study are discussed.
5.2 CFD modelling of cascades with staggered blades

5.2.1 Cascades with aerofoils staggered at 0º
One of the key goals in the present research was to use CFD methods to determine
interference factors for cascades of relevance to OWC axial flow turbines under realistic
flow conditions, i.e. assuming viscous, turbulent flow. A particular interest was the
verification of Weinig’s relationship (Eq. 5.2) for a 2D linear and infinite cascade of
NACA0012 aerofoils (with a stagger angle of 0º) with CFD simulations using the
ANSYS CFX-11 code. All CFD results have been obtained on a PC with hardware
described earlier. Prior to CFD simulations through a range of angles of incidence, grid
independence was verified for each computational domain using a model of a linear
cascade with dimensions of c and s and periodic boundaries as shown in Figure 5.7. The
optimum number of unstructured elements generated for each model of a cascade was
approximately equal to ~1.5 × 106. For all simulations reported in this section the CFX
solver was run using the k-ε turbulence model at Re = 7×105 with the “high resolution”
advection scheme and “auto timescale”. The average cpu time for each run was about 1
hour.
One of the issues that was discovered by the present author during this work was the
importance of employing the appropriate definition of the angle of attack on a cascade
when deducing the lift, drag, axial and tangential force coefficients. In all CFD results
presented below, the effective angle of attack for a cascade was based on the mean of the
upstream and downstream angles of incidence, αm = (α1 + α2)/2 as shown in Figure 5.4
and reported by Weinig (1964).
136
Chapter 5
Periodic
boundaries
Figure 5.7 CFD prediction of streamlines in a linear and infinite cascade of NACA0012
aerofoils (γ = 0º) with s/c=2 and αm = 20º.
The CFD results of interference factor as a function of the inverse of solidity, s/c, and
the mean angle of incidence, αm, are presented in Figure 5.8. The calculated values of a
cascade interference factor, k0 shown in Figure 5.8 (solid line) is given in Table 5.1. The
calculation of k0 was carried out by keeping the chord length, c, at constant value of 0.1m
and varying only the values of spacing, s from 0.125m to 0.33m. This manipulation
produced cascade interference factors for the range of cascade solidities from 0.3 to 0.8
(see, Table 5.1).
Table 5.1 The calculated results of k0 based on Weinig’s relationship, Eq. (5.2).
s/c 1.25 1.42 1.66 2.00 2.50 3.30
solidity 0.8 0.7 0.6 0.5 0.4 0.3
k0 2.45 1.80 1.47 1.27 1.16 1.08
137
Chapter 5
α m = 20ο
4
ko 3
α m = 5ο,10ο
1
α m = 15ο Weinig, (5.2)
0
1.0 1.5 2.0 2.5 3.0 3.5
s/c
Figure 5.8 Lift coefficient interference factor, k0, for a linear cascade (γ = 0º) of
based on Weinig’s inviscid flow analysis, Eq. (5.2).
It is seen from Figure 5.8 that there is generally good agreement between (5.2) and
the CFD results for small angles of attack. However, the agreement is not good for
αm > 10º as a result of the inviscid flow theory not predicting the onset of stall, i.e. the
CFD results show lower values of k0 than Weinig’s results due to stall and the formation
of a separation region on the suction side of the aerofoil which is not accounted for in
Weinig’s inviscid flow analysis (Figure 5.7). However for high solidity cascades (low
s/c) the interference effect between adjacent blades suppresses this separation region
resulting in a steep rise in cascade lift coefficient and hence an increased interference
factor. The blockage effect on the pressure side of an aerofoil in a cascade for two values
of inverse solidity: a) s/c=1.25 and b) s/c=2 at αm = 20º is demonstrated in Figure 5.9.
For blades of finite thickness, small values of s/c lead to an increase in velocity in the
passages and an increase in lift.
138
Chapter 5
Complementing the results presented in Figure 5.8, the drag interference factor, δ0, as
a function of inverse solidity, s/c, for different mean angles of attack is given in Figure
5.10. It can be seen that the value of δ0 remains relatively constant throughout the
investigated range of s/c for small angles, but increases considerably for higher angles of
incidence and this trend sharply rises as the value of inverse solidity decreases.
a) b)
Figure 5.9 Pressure distribution for: a) s/c=1.25 and b) s/c=2 at αm = 20º.
13
10
7
δο α m = 20ο
α m = 15ο
4
α m = 5ο,10ο
1
0.5 1.0 1.5 2.0 2.5 3.0 3.5
-2
s/c
Figure 5.10 Interference factor for drag, δ0, deduced from CFD simulations
complementing the results shown in Figure 5.8.
139
Chapter 5
The cascade interference factor for a linear cascade of NACA0021 aerofoils was also
modelled and validated against the Weinig’s inviscid flow analysis based on Eq.(5.2).
Computed values of cascade interference, k0 are given in Table 5.1.
The boundary settings and applied turbulence model (k-ε) were the same as in the
previous case except the flow conditions. All CFD simulations were carried out at
Re = 2.5×105. It is seen from Figure 5.10 that the CFD results demonstrated again a
generally good match to Weinig’s results (solid line with * symbols) based on (5.2) for
small angles of attack (αm = 5º and αm = 10º). It can be seen also that the results based on
the angle of incidence equal to 16º do not exhibit the stall through the whole range of s/c
compared to results of NACA0012 for αm = 15º presented in Figure 5.8. Moreover, it is
interesting to note that only the results for NACA0021 based on the angle of incidence,
αm = 20º shows the stall at the same value of s/c=1.42 as the CFD data for NACA0012 at
αm = 15º. This confirms the effect of the aerofoil thickness ratio, τ, (Raghunathan, 1995)
and demonstrates that the thicker aerofoils in a linear cascade are able to delay the onset
of stall and flow separation to higher angles of incidence.
3
α m = 5o
Weinig, (5.2)
2
α m = 10o
ko
α m = 20o α m = 16o
0
1.0 1.5 2.0 2.5 3.0 3.5
s/c
Figure 5.11 Lift coefficient interference factor (k0) for a linear cascade (γ = 0º) of
based on Weinig’s inviscid flow analysis, Eq. (5.2).
140
Chapter 5
As mentioned earlier Weinig’s inviscid analysis predicts that the lift interference
factor, k0, for an infinite cascade is independent of the angle of attack, α, and is only a
function of the stagger angle, γ, and ratio of s/c (inverse solidity). CFD analysis of the
linear cascade of practical aerofoils such as NACA00xx series showed that the
interference factors, k0 and δ0, (Figures 5.8 and 5.10) do indeed depend on the angle of
incidence and therefore it may be unwise to rely too heavily on Weinig’s inviscid model
and equations such as (5.2) to predict the lift of blades in a cascade at high angles of
incidence. Nevertheless, Weinig’s relationship (5.2) is applicable to prediction of cascade
interference factor, k0 for NACA0012 and NACA0021 aerofoils arranged in linear
cascades for angles of attack, α, less than 10º. Illustrating this conclusion, Figure 5.11
demonstrates the lift curve (solid line with opened symbols) based on CFD prediction of
cascade flow through the linear cascade of NACA0012 aerofoils at Re = 7×105. CFD
cascade model had the same dimensions (c and s) as were used for correction of cascade
data shown in Figure 5.3. It is evident that CFD prediction is in good agreement with lift
data corrected by applying Weinig’s equation (5.2) for α ≤ 10º. However, the CFD results
exhibit the onset of stall at higher angle of incidence compared to the lift data corrected
by k0.
2
CFD cascade lift data
NACA0012
Cascade lift data
1.5
NACA0012, Eq.(5.2)
CL
1
0.5
Isolated lift data NACA0012
(Sheldahl and Klimas)
0
0 5 10 15 20 25 30
α ο
Figure 5.12 CFD prediction of cascade lift data (solid line with opened symbols),
complementing the results shown in Figure 5.3 for s/c = 1.69.
141
Chapter 5
One of the aims in the present study was to verify the analytical prediction of cascade
interference based on the method of singularities by means of CFD. In order to carry out
this task the results reported by Wong (1994) were used. Wong’s predictions of
normalised, CL, and, Cx, coefficients of NACA0012 aerofoils in a linear cascade through
the range of solidities were presented in the previous section (Figures 5.5 and 5.6).
4
α m = 10ο
3 α m = 5ο
C L /C Lo
Wong (Martensen Method)
2
1
ο α m = 15ο
α m = 20
0
0.2 0.4 0.6 0.8
σ
Figure 5.13 Normalised CL for a linear cascade (γ = 0º) of NACA0012 aerofoils as a

function of solidity, σ =c/s, and the mean angle of incidence, αm: □ 5º; ∆ 10º; ◊ 15º; ○
20º. Broken lines represent CFD prediction; solid line with *symbols is based on Wong’s
inviscid flow analysis (Martensen method).
All CFD data used further for the purpose of comparison were deduced from the same
cascade flow simulations as for verification of Weinig’s results shown in Figure 5.8. A
comparison between normalised, CL, predicted by Wong (1994) and CFD results obtained
through the comparable range of solidities is presented in Figure 5.13. It is clearly seen
that the CFD results produced by cascade flow simulations at mean angles of attack, αm,
up to 15° are in good agreement to results reported by Wong (1994) for solidities less
than 0.5. Further increase in solidity affected the cascade viscous flow presented by CFD
data that exhibit the slightly smaller values of normalised CL for angles αm= 5° and αm=
10° compared to predicted values. On the other hand, CFD results for αm= 15° and αm=
20° demonstrate the similar trend as shown in Figure 5.8 which is to be expected since
142
Chapter 5
the inviscid method of singularities (Martensen Method) cannot model flow separation
and onset of stall typical of real viscous flows through linear cascades at high angles of
attack.
A comparison between a ratio of the axial force coefficient, Cx/Cxo, based on
Martensen method (Figure 5.6) predicted by Wong (1994) and CFD results is presented
in Figure 5.14. Experimental data for αm= 5° (Raghunathan et al., 1985) presented in
Figure 5.6 are also included for comparison. It can be seen that CFD prediction of
normalised Cx for small angles of incidence is in a good agreement with Wong’s
prediction when the values of solidity are less than 0.4. On the other hand, CFD results
for small angles are well matched to experimental data at high solidities (more than 0.6).
CFD prediction of a ratio, Cx/Cxo for high angles of incidence exhibits the similar trend as
depicted in Figures 5.8 and 5.13
5
α m = 20ο
α m = 5ο,10ο
3
C x /C xo Wong (Martensen method)
2
α m = 15ο
1
Experiment (Raghunathan et al. 1990)
0
0.2 0.4 0.6 0.8 1.0
σ
Figure 5.14 Normalised Cx for a linear cascade (γ = 0º) of NACA0012 aerofoils as a

function of solidity, σ = c/s, and the mean angle of incidence, αm: □ 5º, ∆ 10º, ◊ 15º, ○
20º. Broken lines represent CFD prediction; solid line with * symbols is based on Wong’s
inviscid flow analysis (Martensen method); solid line with × symbols represents the
experimental data for αm= 5° (Raghunathan et al., 1985).
As mentioned previously wind tunnel testing of aerofoils in cascades to determine k0

is extremely difficult. Only one set of experimental data showing the mutual interference
143
Chapter 5
between NACA0021 aerofoils in linear cascades of three and five blades reported by
Raghunathan (1988) was found in the public domain literature. One of the conclusions
from that study was that the lift coefficient increases with increasing numbers of blades in
a cascade. In order to verify Raghunathan’s results, the present author simulated an
infinite linear cascade of NACA0021 aerofoils staggered at 0º (using the same Reynolds
number of 2.5×105 and solidity of c/s = 0.5 as Raghunathan’s, see Figures 5.15 and 5.16).
The total number of unstructured elements in the mesh was ~ 2.3 × 106. It is clear from
these graphs that the CFD simulations produced results (broken lines with opened
symbols) that are consistent with the experimental data of Raghunathan, although there
are some differences, which may be attributed to the significant blocking effect that may
have occurred in his wind tunnel experiments.
CFD infinite cascade

2.5
2.0 5 airfoils cascade

(Raghunathan, 1988)
1.5
CL
1.0
3 airfoils cascade
(Raghunathan, 1988)
0.5
0.0
0 20 40 60
αο
Figure 5.15 Comparison of experimental data of CL for cascades of three and five
for an infinite cascade (c/s = 0.5).
144
Chapter 5
CFD infinite cascade
CD
5 airfoils cascade
2 (Raghunathan, 1988)
3 airfoils cascade
(Raghunathan, 1988)
0
0 20 40 60
αο
Figure 5.16 Comparison of experimental data of CD for cascades of three and five
for an infinite cascade (c/s = 0.5).
The results shown in Figures 5.15 and 5.16 provide further evidence as to the
effectiveness of the CFD technique as a means of modelling the behaviour of linear
cascades.
Curran et al., (1998a) published lift and drag characteristics for blades in a
monoplane Wells turbine having a NACA0012 blade profile, rotor solidity of σ = 0.64
and for a Reynolds number of Re = 5.5×105. These lift and drag data appear to have been
estimated from the turbine performance such that any swirl at the rotor was not taken into
account (although the method by which the raw experimental data was processed to give
CL and CD was not entirely clear). The present author has carried out 2D CFD simulations
of airflow in a 2D linear rotor with the same dimensions and Reynolds number. The
unstructured mesh used a total of ~1.6 × 106 elements and the cpu time for each run was
between 35 to 55 minutes depending on the flow angle. The CFD and experimental lift
and drag results are compared in Figures 5.17 and 5.18 together with isolated aerofoil
lift/drag data for the NACA0012 profile (Sheldahl and Klimas, 1981). It is clearly seen
that the CFD lift results are in good agreement with lift values from Curran et al., (1998a)
up to the onset of the stall predicted by the CFD analysis. The CFD analysis shows that
145
Chapter 5
stall occurs at a mean-span angle of attack of approximately 15º. This is opposed to a stall
angle of 12º for the isolated airfoil. However, the results of Curran et al., (1998a) do not
demonstrate such a distinct onset of stall as a CL curve based on CFD data.
3 Curran et al. (1998a)
CL
2D CFD (cascade)
1
Single lift data

(Sheldahl and Klimas, 1981)
0
0 5 10 15 20
αο
Figure 5.17 Comparison of lift coefficients deduced from CFD analysis and experimental
rotor tests (Curran et al., 1998a) for a monoplane Well turbine with NACA0012 blade
profile (σ = 0.64). Isolated aerofoil lift data are also provided (Sheldahl and Klimas,
1981).
In Figure 5.18 the rotor drag coefficients obtained from the CFD analysis are shown
to be lower than those reported by Curran et al., (1998a). This less favourable agreement
can be attributed to the fact that the 2D CFD results do not include effects of
aerodynamic and mechanical losses or the tip clearance effects.
146
Chapter 5
0.9
Curran et al . (1998a)
0.6
CD
2D CFD (cascade)
0.3 Single lift data
0.0
0 5 10 15 20
αο
Figure 5.18 Comparison of drag coefficients (for same conditions as in Figure 5.17).
5.2.2 Aerofoils in a cascades with non-zero stagger angle

A major aim of this phase of the study was to compare, for the first time, the values of
k0 predicted from Weinig’s inviscid flow analysis (as shown in Figure 5.2) to results
deduced from CFD simulations of linear cascades with blades having non-zero stagger
angles. To fulfil this task the full viscous flow CFD analyses were conducted on a
cascade of blades with the same dimensions and inverse solidity (s/c = 1.296) as the 1/3rd
scale model of the Denniss-Auld turbine described by Finnigan and Auld (2003). The
blade profile (see Figure 2.45) used for this prototype turbine was based on the NACA
65-418 profile with maximum camber height of 6% and maximum thickness to chord
ratio of 18%. The blade geometry was symmetric about the mid-chord and was formed
by combining two front halves of the NACA 65-418. Values of lift interference factor k0
deduced from the CFD simulations for a constant upstream angle of attack α = (β -γ) =
10º as a function of stagger angle and solidity are shown in Figure 5.19. The results for
the drag interference factor δ0 = CD/CD0 for the same cascade are also shown in Figure
5.20.
147
Chapter 5
4
γ ′= 70o
3
γ ′= 60o
γ ′= 80o
γ ′= 40o
ko 2
γ ′= 30o
1
γ ′= 20o
γ ′= 0o
0
0 1 2 3 4
s/c
Figure 5.19 CFD prediction of lift interference factor for linear cascade of blades similar
to those of (Finnigan and Auld, 2003) for an upstream angle of incidence α = 10º. Note in
this figure γ′ is measured from the plane normal to the plane of rotation.
8
γ ′ = 40o
7
γ ′ = 30ο
6
δο 4
γ ′ = 70o
3
γ ′ = 0o
2
1
γ ′ = 20o γ ′ = 60o γ ′ = 80o
0
0 0.5 1 1.5 2 2.5 3 3.5 4
s/c
Figure 5.20 Interference factor for drag, δ0, complementing the results shown in Figure
5.19. Note in this figure γ′ is measured from the plane normal to the plane of rotation.
148
Chapter 5
Two-dimensional CFD simulations have been carried out by using an unstructured

mesh having a total number of elements of between 1.2 to 1.45 million depending on the
stagger angle. Cascade flow was modelled at Re=6×105 based on the chord length and the
k-ε turbulence model was used. The average cpu time for each run was about 55 minutes.
Note that the lift and drag coefficients for the cascades have been calculated using the
mean of the upstream and downstream angles of incidence as defined by Weinig (1964).
The CFD results for k0 in Figure 5.19 show a close similarity to the Weinig inviscid flow
analysis results of Figure 5.2 although the magnitudes of k0 predicted from the CFD
analysis are somewhat larger than Weinig’s results (for s/c < 1.5) possibly due to the
increase in velocity around the aerofoils due to the blockage effect from the finite
thickness of the practical aerofoils. Note that the latter also leads to a lower limit to the
value of s/c that can be implemented with practical aerofoils.
For the purpose of direct comparison, Figure 5.21 gives a closer look at Weinig’s data
(broken lines) and CFD results (solid lines) for the stagger angles of 0° and 70°. It is
clearly seen that there is deference between results based on inviscid flow analysis and
results deduced from the fully viscous CFD flow simulations. The difference becomes
more noticeable for the stagger angle of 70° with decreasing the s/c ratio.
CFD, γ ′= 70°
3
ko 2
Weinig, γ ′= 70°
1
Weinig, γ ′= 0°
CFD, γ ′= 0°
0
0 1 2 3 4
s/c
Figure 5.21 Comparison between Weinig’s data (broken lines) and CFD results (solid
lines) for aerofoils staggered at 0° and 70°.
149
Chapter 5
The results for the drag interference factor as a function of stagger angle and inverse
solidity shown in Figure 5.20, demonstrate a trend of increasing interference factor, δ0,
with decreasing spacing of the blades. This behaviour is to be expected. It should be
noted that this increase is not only the result of the effect of the neighbouring blades but
also that increasing solidity results in increased restriction of the axial flow of fluid
through the cascade and thus pressure drop (and hence effective drag) increases rapidly as
the width of the passages between the blades fall below one chord length. Values of k0
and δ0 deduced from the CFD simulations are given in Appendix A-6.
A further comparison, using different cascade geometry, of the cascade interference
factor predicted from Weinig’s theory with those of a fully viscous flow CFD analysis
has been conducted on a cascade of staggered blades with the same dimensions and
inverse solidity (s/c = 0.9) as the full scale model of the Denniss-Auld turbine introduced
in Chapter 2. However, unlike the previous case, where comparison was based on
Weinig’s inviscid flow analysis of the cascade with staggered flat plates (see Figure 5.2),
the comparison given below is based on Weinig’s results of a cascade interference factor,
k1 for circular arc profile shown in Figure 5.22. The main reason for this was the shape of
the aerofoil used in the full scale Denniss-Auld turbine, which closer to the arc geometry
rather than to thin flat plate (see Figure 4.21).
Figure 5.22 Cascade interference factor, k1 for circular arc profile with smooth inflow
conditions (Weinig, 1964). Note in this figure γi is measured from the plane normal to the
plane of rotation.
150
Chapter 5
The full scale Denniss-Auld turbine with 21 cambered blades (c = 230mm) had
solidity, σ = 1.1. Converting this value to the ratio, s/c, gave value of 0.9, which was
employed for a CFD cascade model. Values of lift interference factor k1 deduced from
the CFD simulations for a constant upstream angle of attack α = 0º as a function of
stagger angle and solidity are shown in Figure 5.23 by symbols: □ 20º, ∆ 30º, ◊ 35º, ○
40º, and * 45º. Two-dimensional CFD modelling has been carried out by using an
unstructured mesh having a total number of elements between 1.3 to 1.5 million
depending on the stagger angle. Cascade flow was modelled at Re=1×106 using the k-ε
turbulence model. The average cpu time for each run was about 50 minutes.
In addition to a cascade model based on solidity of the full scale Denniss-Auld
turbine, a linear cascade of aerofoils with the same profile and chord length but a wider
spacing has been modelled. Inverse solidity, s/c, of a new cascade arrangement was set as
1.5. CFD simulations have been conducted using the same Reynolds number (Re=1×106)
and settings of boundary conditions as in a cascade model with smaller spacing (s/c
=0.9). Values of lift interference factor, k1, deduced from the CFD simulations for a
constant upstream angle of attack α = 0º as a function of stagger angle and solidity are
depicted in Figure 5.23 by symbols: □ 20º, ∆ 30º, ◊ 35º, ○ 40º, ∗ 45º, +50º, × 60º, ▲ 75º,
and ■ 80º.
It can be seen that the CFD data for k1 in Figure 5.23 match surprisingly good to the
Weinig inviscid flow analysis results (Figure 5.22) for both cascade models. One of the
possible explanations of this fact is that the shape of aerofoils resembles the circular arc
profile. Secondly, the full viscous CFD simulations were conducted at a constant
upstream angle of attack, α = 0º that provided smooth inflow conditions without flow
separations.
151
Chapter 5
2.0 o
80
o
90
o
70
1.5
o
60
k 1 1.0
γ ′ = 0o
o
o 20
30
0.5 o
50
o
40
0.0
0 1 2 3 4
s/c
Figure 5.23 Comparison between Weinig’s data shown in Figure 5.22 and CFD results
for two linear cascades of the full scale Denniss-Auld aerofoils for an upstream angle of
incidence α = 0º. CFD results for s/c = 0.9: □ 20º, ∆ 30º, ◊ 35º, ○ 40º, * 45º; CFD results
for s/c = 1.5: □ 20º, ∆ 30º, ◊ 35º, ○ 40º, * 45º, +50º, × 60º, ▲ 75º, ■ 80º. Note in this
figureγ′ is measured from the plane normal to the plane of rotation.
The results shown in this section give further evidence regarding efficiency of the
CFD technique as a means of modelling the behaviour of linear cascades with blades
fixed at stagger angles from 0° to 90°.
5.2.2.1 Effect of stagger angle on cascade lift and drag characteristics

One of the aims of study was to clarify the effect of stagger angle on cascade lift and
drag characteristics. In order to perform this task, full viscous flows through a linear
cascade of staggered NACA0012 aerofoils with the same chord length (c =0.055m) and
spacing (s = 0.144m) as the small scale model of the Wells variable-pitch turbine
described by Tease (2003) have been modelled. Two-dimensional CFD simulations have
been carried out through the same range of stagger angles (from 0º to 32º) and at the
same flow conditions (3500RPM) as reported by Tease (2003). In order to achieve the
flow conditions with constant rotational speed of 3500RPM at average blade radius, Rav,
the inlet velocity has been set by applying Cartesian coordinates. The axial component of
flow velocity, VA, has been varied while the tangential component, VT, has been kept as
86.4m/s. The number of unstructured grids used for the CFD modelling of cascade flows
152
Chapter 5
has been varied from 1.4×106 to 1.6×106 depending on the stagger angle. The scalable k-ε
turbulence model has been applied for each flow simulation.
3
3°
11°
0°
2 24°
γ = 32°
CL
Isolated lift data

0
0 10 20 30 40 50
αο
Figure 5.24 Effect of stagger angle on cascade lift data (c/s = 0.49). Isolated airfoil lift
Klimas, 1981). Note in this figureγ is measured from the plane of rotation.
Values of cascade lift and drag coefficients deduced from the CFD simulations as a
function of stagger angle, γ, and angle of attack, α, are presented in Figures 5.24 and
5.25, respectively.
153
Chapter 5
2.4 11°
3°
0°
1.6 24°
CD γ = 32°
0.8
Isolated drag data

(Sheldahl and Klimas,1981)
0.0
0 10 20 30 40 50
αο
Figure 5.25 Effect of stagger angle on cascade lift data (c/s = 0.49) Isolated airfoil lift
Klimas, 1981). Note in this figureγ is measured from the plane of rotation.
It can be seen from Figure 5.24 that the increase of stagger angle shifts the onset of
stall at high angles of incidence to the right, but decreases the maximum values of
cascade lift coefficients. On the other hand, a noticeable reduction in values of cascade
drag coefficients (Figure 5.25) is seen at high angles of incidence with increasing γ.
Similar behaviour of linear cascades of NACA0012 aerofoils fixed at stagger angles of
10º and 30º is reported by Yilbas et al., (1998).
Shifting the onset of stall to higher angles of incidence undoubtedly plays a
favourable role in increasing the operational range of the Wells turbine. This can be
achieved by using variable pitch blades that can be adjusted in response to the change of
the flow conditions through the turbine rotor. Despite the fact that such a variable rotor
configuration increases the complexity of the turbine design; this approach has proved its
effectiveness in several cases reported over the last years (Tease, 2003; Finnigan and
Auld, 2003), at least at laboratory scale.

Theoretical cascade lift interference factors for linear cascades of aerofoils fixed at
stagger angle of 0° predicted by Weinig (1964) and Wong (1994) who used potential
154
Chapter 5
flow methods have been compared in this chapter with a viscous CFD analysis. The CFD
analysis confirmed the applicability of Weinig’s relationship (5.2) for prediction of a
cascade lift interference factor, k0, for the linear cascades of NACA0012 and NACA0021
aerofoils at angles of attack less than 10°.
CFD analysis has also clarified the statement of Weinig’s theory that the cascade lift
interference factor, k0, for an infinite cascade is independent of the angle of attack, α, and
is only a function of the stagger angle, γ, and ratio of s/c (inverse solidity). It was found
that the cascade interference factors are virtually independent of α for small angles, but
equations such as (5.2) must be used with caution to predict the lift of blades in a cascade
at high angles of incidence (αm > 10º).
Comparison between analytical cascade interference data based on the method of
singularities (Wong, 1994) and results deduced from CFD analysis has also shown
relatively good agreement for small angles of incidence (αm < 10º).
The CFD results for k0 obtained by simulations of full viscous flows through the
linear cascade of staggered aerofoils similar to those described by Finnigan and Auld
(2003) showed a close similarity to the Weinig inviscid flow analysis results predicted for
the same range of the stagger angles. However, the magnitudes of k0 predicted from the
CFD analysis were somewhat larger than Weinig’s results (for s/c < 1.5) possibly due to
the increase in velocity around the aerofoils due to the blockage effect from the finite
thickness of the practical aerofoils.
The cascade lift interference factors deduced from CFD simulations of viscous flows
through two cascades (s/c = 0.9, s/c = 1.5) of staggered aerofoils with the same profile as
the full scale Denniss-Auld blades at a constant upstream angle of attack, α = 0º,
demonstrated a good match to Weinig’s data for cascades of aerofoils with circular arc
profiles.
CFD simulations produced cascade lift and data coefficients for the linear and infinite
cascade of NACA002 aerofoils fixed at γ = 0º, that were consistent with the experimental
data of Raghunathan (1998).
CFD prediction of lift characteristics of a monoplane Wells turbine was in good
agreement with lift values from (Curran et al., 1998a) up to the onset of stall, which was
also predicted reasonably well by the CFD analysis.
The effect of stagger angle on cascade lift and drag characteristics has been studied
by means of CFD modelling of a linear cascade of NACA0012 aerofoils with the same
155
Chapter 5
chord length and pitch as the small scale model of the Wells variable-pitch turbine
described by Tease (2003). The investigation confirmed results reported by Yilbas et al.,
(1998) that when the stagger angle was increased the onset of stall moved to higher
angles of incidence. Reduction of cascade drag coefficients at higher stagger angles was
also observed in the CFD results.
156
Chapter 6
CHAPTER 6 - DEVELOPMENT OF A NON-DIMENSIONAL

FORM OF THE BEM ANALYSIS
In this chapter, a non-dimensional BEM model applicable to analysis of axial flow

OWC turbines with fixed and variable rotor geometries is formulated. In addition to this
the application of the developed non-dimensional multiple streamtube BEM model to
predict key performance characteristics of OWC turbines is described.
6.1 Non-dimensional multiple streamtube BEM analysis

Previous actuator disc/blade element turbine analyses reported in the literature have
been implemented using a mixture of dimensional and non-dimensional input parameters.
Results have often been presented in terms of non-dimensional terms, such as flow factor,
φ, efficiency, η, and torque coefficient, CT, yet the underlying analysis has been carried
out using dimensional variables applicable to a particular physical turbine configuration.
One of the issues facing a designer in optimising the configuration of an air turbine
for a particular OWC chamber is that a large number of performance calculations are
required to gain the “damping” (i.e. turbine pressure drop vs. flow rate) required for
maximum hydrodynamic performance of the OWC. In the case of a turbine with variable
rotor geometry the problem is made more complex by the fact that blade stagger angle is
an additional variable to be dealt with. CFD analysis of the air turbine is perfectly
feasible but is too costly in terms of the time required to run a full optimisation. Thus,
blade element analysis of the turbine can be used to get close to the optimal design before
a final CFD analysis is completed.
6.1.1 Non-dimensional BEM model for analysis of OWC turbines

The aim of this part of the chapter is to improve the conventional turbine analysis
based on the actuator disc/blade-element methodology by converting the turbine
coefficients (3.10, 3.11, 3.12 and 3.13) into a purely non-dimensional form. These
manipulations remove many of the limitations of the dimensional analysis and allow
broad comparison of turbine performances using key fundamental non-dimensional
variables. One possible outcome of non-dimensional BEM analysis is the determination
of the optimum solidity, which can be further transformed into a preferred geometry of
the OWC turbine. In subsequent sub-sections the translation of the key turbine
coefficients into non-dimensional groups is performed using the fundamental non-
157
Chapter 6
dimensional parameters that determine turbine performance, i.e. the flow factor, φ,
solidity, σ, blade shape and Reynolds number, Re.
6.1.1.1 Tangential induction factor
WR2 N c (C L sin β − C D cos β )

a' = (3.8)
4π R AV VA Ω (RT + RH )
In order to obtain a non-dimensional expression for a′ , it is necessary to rearrange
(3.8) in the following way. Firstly, variables: N, c, RT and RH are expressed via a
coefficient of solidity, σ, (2.23).
Nc
σ = . (2.23)
π (RT + RH )
Secondly, the expression (CL sin β - CD cos β) in (3.8) can be written as a blade
tangential coefficient Cθ.
Cθ = C L sin β − C D cos β , (6.1)
where β = tan -1((φ /(1−a′))
Thirdly, the resultant relative velocity acting on the blade is given in (3.8):
2
W R2 = V A2 + (Ω R AV (1 + a ′ )) . (6.2)
Dividing the right and the left parts of the (3.2) by (ΩRAV)2 produces (6.3):
2 2
 WR   VA 

 ΩR
 =
  ΩR  + (1 + a ′ )2 . (6.3)
 AV   AV 
VA
Equation (3.14) gives the expression of the flow coefficient as: φ = . Inserting
ΩR AV
(3.14) into (6.3) gives:
2
 WR 
  = φ 2 + (1 + a ′ )2 . (6.4)
 ΩR 
 AV 
Multiplying both right and left parts of (6.4) by (ΩRAV)2 :
( 2
)
WR2 = φ 2 + (1 + a′) × (ΩRAV ) .
2
(6.5)
158
Chapter 6
1 ΩR AV
Keeping in mind that = and finally inserting (2.23), (6.1), and (6.5) into
φ VA
(3.8) gives a non-dimensional expression for a' :
a' =
(φ 2
)
+ (1 + a' )2 × σ × Cθ
. (6.6)
4φ
6.1.1.2 Pressure coefficient

0.5 ρ a WR2 N c (C L cos β + C D sin β )
∆P = (6.9)
π (RT + RH )
Transformation of P* (3.10) into more convenient dimensionless form is attained by
applying the following steps. Firstly, (3.9) can be simplified as follows:
∆P = 0.5 ρ aWR2σC x , (6.7)
where, Cx is the blade axial coefficient and σ is solidity.
Cx = (CL cos β + CD sin β ) (6.8)

Secondly, inserting (6.7) into (3.10), gives:
* WR2σC x
P = 2 . (6.9)
2Ω 2 RAV
Thirdly, inserting (6.5) into (6.9), produces the final form of P*.
σ
P* =
2
[φ 2 2
]
+ (1 + a′) C x (6.10)
6.1.1.3 Input coefficient
∆PQ
CA = (3.11)
(
ρ a V A2 + [ΩRAV ]2 bcN ) VA
2
In (3.11) Q = V A × π ( RT2 − RH2 ) . Inserting (2.23), (6.7) into (3.11) gives:
0.5 ρ aWR2C xπ (RT2 − RH2 )Nc 0.5 ρ aWR2C xπ (RT − RH )(RT + RH )

CA = =
 VA   VA 
( 2 2
)
 ρ a V A + [ΩR AV ] bcN 2 π (RT + RH ) ( )
 ρ a V A + [ΩRAV ] bcN 2 π (RT + RH )
2 2
Simplifying gives:
159
Chapter 6
WR2C x
CA = (6.11)
( V A2 + [ΩR AV ]
2
)
Inserting (6.5) into (6.11) and dividing the numerator and denominator by (ΩRAV)2
gives the final non-dimensional expression for input coefficient:
CA =
(φ 2
(
+ 1 + a ' Cx ))
.
2
(6.12)
φ2 +1
If both sides of the expression (6.10) divide by σ and multiply by 2 it will transform
as follows:
2P*
σ
[
= φ 2 + (1 + a ' ) C x
2
] (6.13)
Inserting (6.13) into (6.12) gives another non-dimensional form of CA:
2 P*
CA = (6.14)
σ φ2 +1( )
6.1.1.4 Torque coefficient
In order to convert the dimensional torque coefficient (3.12) into dimensionless form
the following is carried out. Firstly, the expression for Torque (3.4) is inserted into (3.12):
CT =
2πR AV2
ρ aV A a′Ω RT2 − RH2
=
( 2
2πR AV )
ρ aV A a ′Ω (RT − RH ) RT + RH ( )
( R
ρ a V A2 + [ΩR AV ]2 bcN AV
2
)
ρ a V A2 + [ΩR AV ]2 bcN AV
R
2
( )
and simplified:
4πR AV V A a′Ω (RT + RH )
CT = (6.15)
( 2
V A2 + [ΩR AV ] Nc )
Dividing the numerator and denominator of equation (6.15) by (ΩRAV)2 gives:
4πφa′(RT + RH )
CT = (6.16)
(φ 2 + 1)Nc
1 π (RT + RH )
Bearing in mind that = and inserting this into (6.16) produces the
σ Nc
final non-dimensional form of CT:
4φa′
CT = (6.17)
( 2
φ +1σ )
160
Chapter 6
6.1.1.5 Efficiency coefficient

Following Thakker et al., (2002), Finnigan and Auld (2003) the turbine efficiency is
expressed as:
TΩ С
η= = T (3.13)
∆PQ C Aφ
Inserting (6.14) and (6.17) into (3.13) gives the final non-dimensional form of η:
2 a′
η= (6.18)
P*
6.1.1.6 Advantages of applying a non-dimensional BEM analysis

The main advantage of the non-dimensional BEM method is that it makes possible
using the lift/drag data of the particular aerofoil and specifying the range of flow
coefficients, φ, to conduct the analysis of the performance of various turbine geometries
by changing just one turbine parameter – solidity, σ.
σ
P* =
2
[φ 2 2
+ (1 + a′) C x ] (6.10)
2 P*
CA = (6.14)
(
σ φ2 +1 )
4φa′
CT = (6.17)
( φ2 +1σ )
2 a′
η= (6.18)
P*
The non-dimensional BEM analysis of turbine performance can be conducted by
employing the consequent steps. First, the designer selects an aerofoil profile and uses
related lift/drag data. In case, if the aerodynamic data are not available they can be
generated by means of 2D CFD modelling. Second, the range of solidity values for
analysis are defined and inserted into the dimensionless model. This permits analysis of
the different turbine geometries specified by solidity values simultaneously. The
parameter φ can be also easily altered. This, in turn, allows choosing a turbine
configuration that demonstrates the highest performance. The values of σ and φ, at which
the turbine shows the most efficient operation, are the optimum performance parameters.
The last step is to translate the optimum values of σ and φ into actual turbine
configuration. For instance, the flow coefficient, φ, being the ratio of VA and ΩRT can be
161
Chapter 6
used to define the maximum tip radius, RT, at which the rotational speed of the turbine
rotor does not exceed the critical Mach number, Mcrt. As soon as the RT is determined,
the solidity value, σ, can be used to specify the optimum number of blades, N, and the
optimum chord length, c. Note: the nacelle size as well as the hub radius, RH, is
commonly determined by the size of electrical generator accommodated in the nacelle.
6.1.2 Non-dimensional multiple streamtube model

The BEM analysis presented at the beginning of the current chapter considers the air
duct as a single streamtube where the forces act at the average radius of the blades. In
reality, the tangential induction factor a′ varies along the blade as the radial distance
changes, which in turn causes the forces acting on the blades to vary accordingly. This
can be overcome by applying a multiple streamtube analysis. The essence of this
approach lies in dividing the blades (actuator disk) on elements using concentric thin-
walled streamtubes as shown in Figures 6.1 and 6.2. Each streamtube has separate lift and
drag force components and a separate value for a′. In order to determine a′, and thus the
lift and drag forces the actuator disc/blade element theory is separately applied for each
streamtube. The total torque and pressure drop of the turbine are calculated by adding the
individual components.
Considering a turbine of rotor radius, RT, and hub radius, RH as shown in Figure 6.1,
let the thickness of each streamtube be dr. For n equally spaced streamtubes, dr = (RT -
RH)/n as well as RT=RTn, RH=RHn, and RAV=RAV,n.
Streamtubes
RH
RT
dr Disc
Figure 6.1 Four streamtubes air duct representation.
162
Chapter 6
Forces
Hub
Figure 6.2 Four streamtubes blades representation.
In the present study, the axial flow turbines were evaluated using the multiple
streamtube analysis based on the non-dimensional model developed in sub-section 6.1.1.
The calculation procedure was as follows. Firstly, the non-dimensional tangential flow
induction factor given in (3.20) was computed for each streamtube as:
a' =
(φ 2
)
+ (1 + an ' ) 2 × σ n × Cθ , n
n , (6.19)
4φ
where n is the subscript referring to the current streamtube and its individual components.
Solidity of the turbine rotor within each streamtube was calculated as:
Nc Nc
σn = = , (6.20)
π (RT ,n + RH ,n ) 2πR AV ,n
where, RAV,n =(RT,n +RH,n)/2 is the average radius of each blade element.
The non-dimensional turbine performance characteristics P*, CA, CT, and η of each
blade element were iteratively calculated using equations (6.10), (6.14), (6.17) and
(6.18):
σn
*
Pn =
2
[φ 2 2
]
+ (1 + a′n ) C x ,n (6.21)
*
2 Pn
C A,n = (6.22)
σ n (φ 2 + 1)
4φa′n
CT , n = (6.23)
(
φ + 1σn 2
)
2 a ′n
ηn = * (6.24)
Pn
Finally, the overall P*, CA, CT, and η were calculated by adding the respective
components for each streamtube such that:
163
Chapter 6
 n 
 ∑ ∆Pn* 
P* =  1 
(6.25)
n
 n 
 ∑ C A,n 
CA =  1 
(6.26)
n
 n 
 ∑ CT ,n 
CT =  1 
(6.27)
n
In the present multiple streamtube analysis only the non-dimensional independent

variables are used to predict the turbine performance characteristics. They were derived
in chapter 3 as follows:
σn
*
Pn =
2
[φ 2 2
]
+ (1 + a′n ) C x ,n (3.35)
*
2 Pn
C A,n = (3.36)
σ n (φ 2 + 1)
4φa′n
CT , n = (3.37)
( 2
φ + 1σn )
2 a n′
ηn = * (3.38)
Pn
 n 
 ∑η n 
η= 1  (6.28)
n
164
Chapter 6
6.2 Analysis of the 1/3rd scale Denniss-Auld turbine

In the present study, the multiple streamtbe non-dimensional model was initially
applied to a laboratory-scale model of a Denniss-Auld turbine. The turbine configuration
was the same as that tested in a wind tunnel by Finnigan and Auld (2003). The overall
solidity of this turbine was equal to 0.77 and major radii: RH = 0.1m, RT = 0.23m. In
applying the multiple streamtube model, the span, b = RT -RH of each turbine blade was
divided into 10 equally spaced strips. The non-dimensional analysis was conducted
through a range of flow coefficients, φ, from 0 to 8. Input lift and drag coefficients
calculated from the previous CFD analysis described in chapter 4 used in the non-
dimensional analysis and are given in Appendix A-2. Numerical analysis was carried out
using MatLab. It should be noted, that numerical results presented below are given
without inclusion of kinetic energy losses from the turbine diffuser, which can be
significant in practical situations. This issue will be discussed in chapter 8.
The predicted efficiency as a function of blade stagger angle, γ, and flow factor, φ, is
shown in Figure 6.3 It is seen that the simulation results based on 10 streamtube model
(broken lines) and experimental data have the same general shape, but the non-
dimensional model overestimated the efficiency at all stagger angles tested in study. The
peaks of the predicted efficiency for stagger angles of 20º and 40º were higher by 19%
and 17% than in the experiments, respectively. For stagger angles of 60º and 80º, the
peaks of efficiency curves were overestimated by 36% and 41%, respectively. The
significant differences between the simulated and experimentally measured efficiency
were possibly due to: the relatively large frictional losses in the laboratory model; the fact
that blade Reynolds numbers were low, and due to other losses (e.g. tip losses) that were
not modelled here using the BEM. According to Finnigan and Auld (2003), the blades of
the 1/3rd scale turbine were not precisely finished and the tip clearance varied from 2mm
to 5mm. Despite correspondence with the authors (Finnigan and Auld, 2003) it was not
possible to recover sufficient details of experimental apparatus and testing methodology
to determine whether frictional losses in the drive train and elsewhere were included in
the calculation their results for efficiency and other coefficients.
165
Chapter 6
0.9
single streamtube
20° 10 streamtubes
40°
0.6
η 60°
0.3
γ = 80°
0
0 2 4 φ 6 8
Figure 6.3 Numerical results (solid lines) from the present 10 streamtube BEM analysis
(broken lines) for the efficiency of a laboratory-scale Denniss-Auld turbine (isolated
aerofoil lift/drag data deduced from CFD simulations) compared with experiments (□
γ =20º, ○ γ = 40º, ∆ γ = 60º, ◊ γ =80º) by Finnigan and Auld (2003). Results based on
single streamtube BEM model (solid lines) are also given for comparison.
Figures 6.3 to 6.6 show a comparison of the experimental results and MatLab
simulation results obtained for torque, CT, input, CA, and pressure, P*, coefficients. The
comparisons show that the non-dimensional model matches the experimental data
relatively well for the input, CA, and pressure, P*, coefficients for all stagger angles
except 80°, at least for flow factors, φ, up to those corresponding to the onset of stall
predicted by the simulations. It is clearly seen, that the input, CA, and pressure, P*,
coefficients for γ = 80° were underestimated approximately by 51% and 46%,
respectively. It is also seen that, the match between the experiments and the blade-
element model for torque coefficient, CT, is not so good and there is a significant
difference between results. This situation can be explained by the inherent limitations in
the BEM model based on isolated lift/drag data to predict accurately the torque
characteristic of turbines having relatively high rotor solidities (σ = 0.77 in the present
case). The numerical results clearly show the predicted effects of blade stall at small
stagger angles for both input, CA, and torque, CT, coefficients, which do match reality due
to lack of consideration of cascade effects. This is especially noticeable for γ = 20º and γ
166
Chapter 6
= 40º, but as soon as the stagger angle increases further together with increasing of the
flow coefficient then stall does not affect the blades to the same extent.
It is interesting to note the very small differences between the results from the single
streamtube model and the 10 streamtube model. The single streamtube model was based
on the average blade radius, RAV = 0.165m. It can be seen from Figures 6.5 – 6.6 that
there is slight difference in the stall region between predicted CT and CA coefficients
based on 10 streamtube (broken lines) and single streamtube (solid lines) models for γ =
20º and γ = 40º as well as between predicted P* for all stagger angles. The most
noticeable difference between numerical results was shown by simulation with γ = 40º,
where the 10 streamtube model predicted more smooth turbine operation in the post-stall
interval.
1.6 single streamtube

10 streamtubes
20°
1.2
CT 40°
0.8
60°
0.4
γ = 80°
0
0 2 4 6 8
φ
Figure 6.4 Torque coefficient, CT, predicted from isolated lift/drag data deduced from
CFD simulations. Note in this figure γ is measured from the plane of rotation.
167
Chapter 6
3
single streamtube
10 streamtubes
2 20°
CA
1 40°
60°
γ = 80°
0
0 2 4 6 8
φ
Figure 6.5 Input coefficient, CA, predicted from isolated lift/drag data deduced from CFD
simulations.
10
single streamtube
10 streamtubes
8
6
P*
4
40°
2 20° γ = 80°
60°
0
0 2 4 6 8
φ
Figure 6.6 Pressure coefficient, P*, predicted from isolated lift/drag data deduced from
CFD simulations.
168
Chapter 6

The non-dimensional BEM analysis based on the 10 streamtube model has been
shown to be useful in predicting the key performance characteristics of OWC turbines
with variable pitch angle blades. However, possibly due to limitation of the BEM
analysis, which did not include the aerodynamic and frictional losses, the performance
characteristics, in terms of η and CT for the small scale Denniss-Auld turbines were
overestimated.
169
Chapter 7
CHAPTER 7 - CFD ANALYSIS OF AXIAL FLOW TURBINES

One of the goals of the present research was to investigate the utility of three
dimensional 3D CFD simulations of axial flow turbines designed to service OWC wave
energy converters. This chapter presents results of 3D CFD modelling of actual OWC
turbine designs. Firstly, the Wells turbine with variable pitch blades was modelled and
analysed by means of a commercial CFD code, ANSYS CFX-11. Secondly, the
performance characteristics of a 1/3rd scale Denniss-Auld turbine were modelled and
compared with available experimental data. Thirdly, the influence of different design
parameters on performance of the full scale Denniss-Auld turbine was investigated and
analysed providing guidance for further turbine optimisation.
7.1 High solidity Wells turbine model with variable pitch blades
In order to conduct 3D simulations of the Wells turbine with variable pitch blades, the
dimensions and operational parameters of an actual turbine reported by Tease (2003)
were used in the present study.
To model airflows through this type of turbine, one of the first and most important
steps was to replicate the blade geometry and to develop a grid with appropriate
topology/resolution, particularly in the vicinity of blade. Both structured and unstructured
meshes were applied to model the turbine performance through the investigated range of
stagger angles. The reason for using these two types of mesh was that it was not possible
to use the TurboGrid code to create structured meshes around blades staggered at angles
of between 0º and 20º relative to the plane of rotation. Thus, the present author used the
standard CFX-Mesh generator to build unstructured grids in combination with inflation
layers to overcome this problem for this range of angles. Based on the grid independence
testing, a total of ~106 elements were used in the structured grids for large stagger angles
(γ = 24º, and γ = 32º). For smaller angles (γ = 3º, γ = 11º) an unstructured mesh was
employed with a total number of elements of ~2.6 × 106. A typical example of a
structured mesh with a blade stagger angle of γ = 32º is illustrated in Figure 3.10. All
CFD results presented in this section were obtained on a PC with hardware listed in
chapter 4. For each 3D CFD run the cpu time taken was about two hours.
170
Chapter 7
The results reported by Tease (2003), as part of the Wavegen variable pitch Wells
turbine research program (details of the actual turbine configuration are given in Table
7.1.), were used for verification of 3D CFD results obtained during the present study.
Table 7.1 Details of the Wells turbine with variable pitch blades reported by Tease
(2003)
Tip radius, RT 0.2775m
Hub-to-tip ratio, h 0.7
Number of blades, N 13
Chord length, c 55mm
Blade profile NACA0012
Tip clearance, tc 1.5mm
Rotational speed, ω 3500 RPM
A great deal of work was carried out prior to 3D CFD analysis of the actual OWC
turbines to determine the best model for the rotational turbine blades and motionless
components such as: air duct and nacelle. A single domain with a rotational frame of
reference was used on initial stage of CFD modelling of the Wells turbine as reported by
Tease (2003). However, this approach produced CFD results, which were in significant
disagreement with experimental data. The possible reason for this is rather complex
arrangement of the turbine test rig. It can be seen from Figure 7.1 that the variable pitch
turbine being a rotational part of the test rig is coupled to the non-rotational nacelle. Take
this into account the present author decided to use ANSYS-CFX Turbo mode rather than
General mode and employ three domains including two stationary and a rotational
domains as depicted in Figure 7.2. The author’s opinion that this approach made possible
to model the actual Wells turbine tested by Tease (2003) more realistically. The domains
(stationary, rotational and stationary) were coupled by general connection interfaces with
a Frozen Rotor frame change. Non-slip walls were specified at the blade surface, hub,
shroud, and nacelle and rotational periodicity was applied along the meridian surfaces for
each of the thirteen blades. The tip clearance, tc, was set to 1.5mm as in the actual turbine.
171
Chapter 7
Figure 7.1 Schematic of the turbine test rig (Tease, 2003).
Inlet: Vnor
Stationary domains
Rotational domain
General connection interfaces
Outlet: Pst
Figure 7.2 Computational domain used for modelling of the Wells turbine reported by
Tease (2003) with thirteen NACA0012 blades staggered at γ = 3º and σ =0.48.
In order to provide the same range of flow coefficients, φ, as in the experiments

reported by Tease (2003), the rotation of the CFD turbine model was kept at a constant
value of 3500RPM while the inlet velocity was varied.
One of the great advantages of 3D CFD analysis is the opportunity it provides to
visualise the important flow features. This in turn, assists in interpretation of numerical
results deduced from CFD simulations. Visualisation of a flow field in the vicinity of a
Wells turbine blade with γ = 3° is shown in Figures 7.3 to 7.4. The values of φ given in
figures: 0.12, 0.29, 0.46 and 0.64 correspond to the values of the mean angles of attack,
172
Chapter 7
αm: 3.6°, 13.6°, 21.9° and 29.5°. Deep stall characterised by significant flow separation
at αm = 21.9° and αm = 29.5° can be seen in Figures 7.3 (c, d) and 7.4 (c, d).
a) b)
φ = 0.12 φ = 0.29
c) d)
φ = 0.46 φ = 0.64
coefficients, φ.
a) b)
φ = 0.12 φ = 0.29
c) d)
φ = 0.46 φ = 0.64
Figure 7.4 Distribution of streamlines in vicinity of a Wells turbine blade with γ = 3° at

different flow coefficients, φ (complementing Figure 7.3).
Numerical results for the non-dimensional pressure coefficient, P*, deduced from 3D
CFD simulations of the Wells turbine model with variable pitch blades are presented in
173
Chapter 7
Figure 7.7. as a function of flow factor, φ, and γ, together with experimental results
reported by Tease (2003). It can be seen that the CFD results demonstrated the same
trend as in experiments; they moved to the higher values of φ with increasing γ and had a
similar slope. CFD analysis revealed that the pressure coefficients, P*, deduced from
CFD simulations carried out at γ = 3º were predicted quite accurately at small values of φ.
However, in contrast to the experimental data, the CFD model predicted the onset of stall
at φ = 0.29 (see, Figures 7.3 (b) 7.4 (b)) and overestimated the measured coefficient, P*,
at this point by 14%. Contrary to this, at φ = 0.46 the CFD model produced a value of P*
on 9% lower than in experiments. It is also seen that the CFD results were predicted
relatively accurately within the interval of 0.46 < φ < 0.64. For the angle of γ = 11º the
CFD model predicted the pressure coefficient, P*, quite well through all values of φ tested
in study. A comparison of measured and predicted pressure coefficients at γ = 24º showed
that the CFD model underestimated P* (by approximately 20%) for 0.52 < φ < 0.78. It is
clearly seen that the P* data were predicted quite accurately for 0.78 < φ < 0.98. Increase
of φ from to 0.98 to 1.08 demonstrated underestimation of the measured P*data by 5%. It
is interestingly to note that the pressure coefficient, P*, measured at γ = 24° exhibited the
onset of the pseudo stall at φ = 0.69 shown by a red circle in Figure 7.7. However,
visualisation of flow field in vicinity of the blade at the same φ did not demonstrate such
flow behaviour. The flow remained attached to the suction side of the blade (see, Figures
7.5 (a) to 7.6 (a)). Only at higher φ = 0.92, the onset of stall and flow separation near the
hub was observed (see, Figures 7.5 (b) to 7.6 (b)).
a) b)
φ = 0.69 φ = 0.92
coefficients, φ.
174
Chapter 7
a) b)
φ = 0.69 φ = 0.92
blade with γ = 24° at different flow coefficients, φ, (complementing Figure 7.5).
1.5
ο γ =32o
3
1.0
11o 24
o
*
P
0.5
0.0
0.0 0.3 0.6 0.9 1.2 1.5 1.8
φ
Figure 7.7 Results from the present 3D CFD simulations (closed symbols/broken lines)
for non-dimensional pressure, P*, as a function of flow factor, φ, for different stagger
angles compared with the experimental results (open symbols) reported by Tease (2003).
175
Chapter 7
Analysis of both sets of data for γ = 32º revealed that the experimental P* data were
underestimated approximately by 4% for 0.81< φ <1.04. At the higher values of φ the P*
coefficients deduced from 3D CFD simulations demonstrated very close agreement to the
measured data. The average difference between both sets of P* for 1.04 < φ < 1.52 was
less than 1%. As in previous case, the onset of the pseudo stall was also indicated by the
measured data. This point at φ = 0.98 is shown by a blue circle in Figure 7.7.
Visualisation of flow field in vicinity of the blade staggered at γ = 32º did not indicate the
onset of stall at φ = 0.98 (see, Figures 7.8 (a) to 7.9 (a)). Even at higher φ = 1.16, the flow
remained attached almost throughout the suction side of the blade (see, Figures 7.8 (b) to
7.9 (b)).
a) b)
φ = 0.98 φ = 1.16
coefficients, φ.
a) b)
φ = 0.98 φ = 1.16
blade with γ = 32° at different flow coefficients, φ, (complementing Figure 7.8).
176
Chapter 7
A plot of torque coefficient, CT, as a function of flow factor, φ, as deduced from 3D

CFD simulations of the Wells turbine of Tease (2003) is shown in Figure 7.10. It is
clearly seen that increasing the blade stagger angle of the Wells turbine leads to an
increase in the flow factor at which the onset of stall occurs and also leads to an increase
in the magnitude of the torque coefficient, CT. This trend can be explained by the ability
of Wells turbine with variable pitch blades to respond effectively to an increase in the
flow factor by adjusting the blade stagger angle in such way that the angle of attack angle
is maintained in the pre-stall region, which provides the most efficient turbine
performance.
0.9
24o γ =32o
0.6
11o
CT
0.3
3o
0.0
0.0 0.3 0.6 0.9 1.2 1.5 1.8
φ
Figure 7.10 Variation of torque coefficient, CT, from 3D CFD simulations by the present
author of a variable pitch angle Wells turbine (Tease, 2003).
7.1.1 Comparison of 3D CFD and blade element analyses

In the present work the 3D CFD results described above were compared with results
obtained by applying the BEM 10-streamtube model for pressure coefficient, P*, and
torque coefficient, CT, with a view to assessing the robustness of the blade element
model.
The isolated lift and drag coefficients, CL and CD, for NACA0012 aerofoil used in the
blade element model were taken from Sheldahl and Klimas (1981). A comparison
177
Chapter 7
between pressure coefficients, P*, based on 3D CFD and BEM data are shown in Figure
7.11. It can be seen that the agreement between all 3D CFD and BEM results is generally
good in pre-stall region, but the BEM analysis predicted deep earlier stall for each stagger
angle compared to CFD model. On the other hand, at stagger angles of 11º, 24º and 32º
the 3D CFD models predicted only mild stall that occurred at φ: 0.52, 1.04 and 1.39,
respectively. However, at stagger angle of 3º the CFD analysis showed a quite noticeable
stall at φ of 0.46 (see, Figures 7.3 (c) to 7.4 (c)). In summary, the 3D CFD analysis
predicted milder stall at higher values of φ compared to the BEM model based on isolated
lift/drag data. The possible explanation of this is that CFD model takes into account the
cascade effect and other three-dimensional flow effects within the real rotor, which delay
the onset of stall and make it milder.
1.5
o
γ =32o
3
1.0 o o
11 24
P*
0.5
0.0
0.0 0.3 0.6 0.9 1.2 1.5 1.8
φ
Figure 7.11 Comparison of pressure coefficient results from 3D CFD simulations
(closed symbols/broken lines) and from the BEM 10-streamtube model (open
symbols/solid lines), for the geometry of the Wells turbine reported by Tease (2003).
Similar data for CT was also obtained using the BEM model. Results of 3D CFD and
BEM analysis are compared in Figure 7.12. It can be seen that the general form of the
results in the pre-stall is qualitatively the same; however there are some quantitative
differences between both sets of CT data. This can be attributed to the limitation of the
BEM model based on the isolated lift/drag coefficients, which did not capture important
178
Chapter 7
3D aspects of the flow in the actual rotor modelled by CFD. It is clearly seen that the
BEM analysis overestimated CT and predicted earlier stall for all stagger angles compared
to the CFD analysis. The best match in the pre-stall region was achieved at γ =3º where
the average difference between BEM and CFD results was within of 1%. At higher
stagger angles of 11º, 24º and 32º, the BEM analysis showed the onset of stall at φ of
0.45, 0.79 and 1.06 whereas the CFD model predicted the onset of stall at φ of 0.52, 0.92
and 1.16 respectively. The average difference between both sets of data for the stagger
angles of 11º, 24º and 32º in the pre-stall region was 11%, 12% and 7%, respectively.
0.9 γ =32o
o
24
0.6
o
11
CT
0.3
o
3
0.0
0.0 0.3 0.6 0.9 1.2 1.5 1.8
φ
Figure 7.12 Comparison of numerical results based on a blade element model (open
symbols/solid lines) and isolated lift/drag data taken from Sheldahl and Klimas (1981)
against results for CT deduced from 3D CFD simulations (closed symbols/broken lines).
Both set of data obtained in present study for a Wells turbine reported by Tease (2003).
In summary, the BEM analysis based on isolated lift/drag data and a 10-streamtube
model predicted the performance characteristics of the Wells turbine with variable pitch
blades (σ = 0.48) relatively accurately as compared with the 3D CFD model. BEM
analysis being a simpler prediction tool than 3D CFD analysis can be used effectively on
the initial design stages to determine the major performance parameters of axial flow
turbines. As soon as these parameters are clarified, the 3D CFD analysis can be applied
to optimise the turbine performance.
179
Chapter 7
7.2 Small scale Denniss-Auld turbine model

This section presents results of 3D CFD simulations of a 1/3rd scale Denniss-Auld
turbine model shown in Figure 4.16. The turbine was characterised by the following
parameters: RH = 100mm, RT = 230mm, c = 100mm and N = 8 and the blades were
symmetrical about mid-chord as shown in Figure 2.45. The CFD results were obtained by
modelling a 1/8th part of the turbine and imposing rotational periodic boundaries along
meridian surfaces. Since the actual turbine had a non-rotational nacelle, the present
author built a computational domain consisting of three parts (stationary, rotational and
stationary) coupled by general connection interfaces (GUI) with a Frozen Rotor frame
change. The domain was discretized by structured elements using the TurboGrid mode.
No-slip boundary conditions were specified at the blade surface, hub, and shroud. To
achieve the same range of flow coefficient, φ, as in the experiments all simulations were
carried out by modelling an actual rotational speed of the rotor and inlet air velocity.
Prior to completing the schedule of 3D CFD simulations, a set of preliminary runs
were performed to determine the effects of varying the inlet turbulence intensity. ANSYS
CFX-11 had three options to model the inlet turbulence intensity: low-1%, medium-5%
and high-10%. It was found that variation of turbulence intensity from 1% to 10% had
negligible effect on the final output. For this reason the default values of 5% intensity
were used. The k-ε turbulence model with scalable wall functions was applied to simulate
airflows through the turbine. A zero static pressure condition was imposed at the outlet.
Since the actual small scale turbine according to Finnigan and Auld (2003) had blades
with tip clearance, tc, which was varied from 2mm to 5mm, the value of tc in all CFD
simulations and for all stagger angles was set as 4mm. The values of torque and pressure
drop were directly deduced from 3D CFD simulations. All CFD results presented in sub-
section 7.2.2 were obtained using a PC.
7.2.1 Grid independence testing

Prior to the main simulations, the computational domains having the number of grid
elements from 0.6 to 1.2 million were tested. The grid independence tests revealed that
the increase in number of elements to more than 0.9×105 changed the numerical results
less than 1%. Based on this fact, the final number of grid elements was set from 0.9×105
to 1.1×106 depending on the stagger angle settings.
180
Chapter 7
7.2.2 Comparison of 3D CFD and experimental results

A comparison of CFD and experimental results for efficiency, η, indicated an
excellent agreement between experiments and results of the 3D CFD modelling for
stagger angles of 20º and 40º (Figure 7.13). However, predictions for η of 60º and 80º
were less favourable. The measured efficiencies for these stagger angles were
overestimated by approximately 18% and 21%, respectively.
0.7
40°
0.6 20°
0.5
η 0.4
0.3 60°
0.2
γ = 80°
0.1
0.0
0 2 4 6 8 10
φ
Figure 7.13 3D CFD results (closed symbols/solid lines) from the present study for the
efficiency of a 1/3rd scale Denniss-Auld turbine model compared with experiments (open
symbols: □ γ =20º, ○ γ = 40º, ∆ γ = 60º, ◊ γ =80º) by Finnigan and Auld (2003). Note in
this figureγ is measured from the plane of rotation.
181
Chapter 7
3.5
2.8
2.1
CT
20°
1.4 40°
0.7
60°
γ = 80°
0.0
0 2 4 6 8 10
φ
Figure 7.14 Predicted torque coefficient, CT, (opened symbols/solid lines) compared with
Auld (2003). Note in this figure γ is measured from the plane of rotation.
3D CFD analysis showed that the 3D CFD model overestimated experimental torque
coefficients, CT, for all stagger angles (Figure 7.14). The most significant difference
(~71%) between the measured and CFD data was shown at γ = 80°. The less significant
difference (by approximately 42%) between the experimental and CFD results was
demonstrated at γ = 40°. All discrepancies between both sets of data shown in Figure 7.14
may be due to the frictional and/or other losses produced in the apparatus employed to
measure the torque in the wind tunnel. As mentioned earlier despite contacting the lead
author (Finnigan and Auld, 2003) it was not possible to determine whether any losses
were included in the calculation their results including the CT data presented here.
It can be seen from Figure 7.15 that the 3D CFD model predicted CA data at γ = 80°
quite accurately, but for other stagger angles the input coefficients were overestimated as
in a case of CT. Average overestimations of CA for γ of 20°, 40° and 60° were 52%, 33%
and 26%, respectively.
182
Chapter 7
12
CA
4 20°
40° γ = 80°
60°
0
0 2 4 6 8 10
φ
Figure 7.15 Predicted input coefficient, CA, closed symbols/solid lines) compared with
Auld (2003). Note in this figureγ is measured from the plane of rotation.
It can be seen from Figure 7.15 that the CFD model predicted CA data at γ = 80° quite
accurately, but for other stagger angles the input coefficients were overestimated as in a
case of CT. Average overestimations of CA for γ of 20°, 40° and 60° were 52%, 33% and
26%, respectively.
7.2.3 Modelling of the full turbine coupled to the nacelle

One of the aims of the present research was to carry out 3D modelling a 1/3rd scale
Denniss-Auld turbine with eight blades coupled to the nacelle. Since such approach from
the point of view of the present author might provide more realistic representation of the
actual turbine tested in the wind tunnel, it was expected that the accuracy to predict the
performance data might also be improved. To achieve this, the full 3D CFD models of a
1/3rd scale Denniss-Auld turbine with eight blades staggered at 20º and 80º coupled to the
non-rotational nacelle were built. Isometric view of one turbine model is shown in Figure
7.16. In order to take into account the non-rotational nacelle, the computational domain
of each turbine model had three sections (S1, R1 and S2) coupled by general connection
interfaces with a Frozen Rotor frame change. The domains were discretized by the
unstructured meshes using CFX-Mesh mode. Based on the grid independence testing, the
183
Chapter 7
overall number of elements was generated to be equal to 3.7×106 for both turbine models.
Five inflated layers were used on each blade to model the boundary layer in the near wall
region. The full turbine models were pre-processed in the ASNYS CFX-Turbo-mode and
the boundary conditions and turbulence model were the same as in the case of modelling
a 1/8th of the turbine. The tip clearance was set as 4mm.
A cluster computer having the following hardware was used to run all simulations:
FrontEnd (Main computer): Intel Xeon E5410@2.33GHz, C2 quad core = 8 cores and
8GB of RAM; Compute nodes: 80GB disc HDD, 2quad core E5410@2.33GHz and 8GB
of RAM. This hardware specification is henceforth referred to as “the cluster computer”.
The performance data deduced from 3D CFD simulations of the full turbine models with
the inclusion of the nacelle and CFD results from sub-section 7.2.2 are presented in
Figures 7.17 to 7.19. CFD analysis of the torque coefficients for both models and both
stagger angles tested in study revealed that torque characteristics are in close agreement
(Figure 7.18). The same can be said regarding the input coefficients for γ = 80º (Figure
7.19). However, less favourable agreement was shown in the case of γ = 20º, where a full
turbine model produced CA coefficients approximately by 11% higher compared to a 1/8th
of the model. A comparison of both sets of data in terms of predicted efficiency showed
that the full turbine model with blades staggered at 20° produced CFD results in average
by 2% lower than a model with one blade (Figure 7.17). The opposite results were
demonstrated at γ = 80º, where a 1/8th of the turbine model predicted η by approximately
2% lower compared to the full turbine model. The discrepancies between performance
data shown by both models can be explained by the fact that the full turbine models had
much coarse mesh than the models with one blade. The average number of structured
elements used in the 1/8th of the turbine model was 1×106 whereas the 1/8th part of the full
turbine had only ∼ (3.7 × 106)/8 = 0.46×105.
In summary, 3D CFD modelling of the full turbine coupled to the nacelle did not
significantly improve the match with performance data but rather confirmed the results
predicted by the model with one blade described in the previous sub-section.
184
Chapter 7
S2
R1
S1
Figure 7.16 Isometric view of a 1/3rd scale Denniss-Auld turbine model with blades
staggered at 20º. The blade rotation is shown by the red arrow.
0.6 γ = 20º
Full turbine
1/8th turbine
0.4
γ = 80º
η
0.2
0.0
0 2 4 6 8 10
φ
Figure 7.17 Comparison between efficiencies, η, deduced from 3D CFD modelling of a

full 1/3rd scale Denniss-Auld turbine (open symbols/solid lines) and a 1/8th of the turbine
(closed symbols/broken lines). Experimental data (symbols: ▲20°,♦80°).
185
Chapter 7
3.5 γ = 20º
Full turbine
1/8th turbine
2.8
2.1
CT
1.4
0.7
γ = 80º
0.0
0 2 4 6 8 10
φ
Figure 7.18 Comparison between torque coefficients, CT, deduced from 3D CFD
of the turbine (closed symbols/broken lines). Experimental data (symbols: ▲20°,♦80°).
12
γ = 20º Full turbine
1/8th turbine
9
CA
6
γ = 80º
0
0 2 4 6 8 10
φ
Figure 7.19 Comparison between input coefficients, CA, deduced from 3D CFD
of the turbine (closed symbols/broken lines). Experimental data (symbols: ▲20°, ◊ 80°).
186
Chapter 7
7.3 Full scale Denniss-Auld turbine model

One of the aims of the present study was to investigate the influence of key design
parameters on performance a full scale Denniss-Auld turbine introduced in Chapter 2. To
achieve this, the turbine model was modelled using the same boundary conditions as in
case of the small scale model described in the previous section. The computational
domain for each run was composed of three parts (stationary-S1, rotational-R1 and
stationary-S2) coupled by general connection interfaces with a Frozen Rotor frame
change. The results of 3D CFD simulations presented in this section were obtained on a
PC and the cluster computer.
Most of the time the turbine model was run with blades staggered at angles of 45º, 55º
and 70º and tc =4.6mm, which was comparable with tip clearance of the actual turbine. A
range of flow conditions comparable with the full-scale tests carried out on the device
with real sea conditions at Port Kembla was attained by keeping the constant angular
velocity of rotational domain (36.6rad/s or 350RPM) and by varying the axial air velocity
at the inlet. The pressure drop data across the turbine rotor were taken directly from 3D
CFD modelling. The turbine shaft torque was computed by finding the average torque on
each of the blades.

In order to determine the optimal number of structured grid elements for the
computational domain, the grid independence tests were carried out before conducting of
the main simulations. The structured elements were generated using TurboGrid and their
number was tested from 0.7×105 to 1.2×106. The tests showed that the optimum number
of elements, which did not substantially affect the numerical results, was 1×106. Based on
this information, the final number of structured elements for a computational domain was
set from 1×106 to 1.1×106 depending on the value of the stagger angle.
7.3.2 Effect of design parameters on turbine performance

For the present study, the following turbine design parameters were investigated to
determine their influence on overall turbine performance and to investigate how best to
optimise the configuration of a Denniss-Auld turbine for a particular application:
• Tip clearance, tc.
• Hub to tip ratio, h.
• Rotor solidity, σ.
187
Chapter 7
• Blade profile.
• Blade thickness ratio, τ.
7.3.2.1 Tip clearance, tc

It should be noted that no data were found in the open literature regarding the effect
of tip clearance, tc, on performance of axial flow variable pitch angle turbines. However,
the direct relation between the value of tc and efficiency of Wells turbine with fixed
blades were reported by Raghunathan (1995), Torressi et al., (2007) and Torressi et al.,
(2008).
In the present study the effect of tc on the performance of the full scale Denniss-Auld
turbine model was investigated at stagger angles of 45º, 55º and 70º. Values of tc, equal to
0mm, 2.3mm, 4.6mm and 6.9mm or 0%,1%, 2% and 3% of the chord length (230mm),
respectively, were tested in the operating conditions typical for these stagger angles. The
tip clearance or tip gap (TG) for all tested values was meshed by structured elements as
shown in Figure 7.20. The increase of value of tc was attained by decreasing the blade
span while the hub and shroud radii were kept unchanged. The tip clearance in all CFD
simulations was to be uniform along the blade chord regardless of the stagger angle, γ.
However, it should be noted that the tip clearance in an actual variable blade pitch turbine
will be both a function of distance from the mid-plane of the aerofoil and stagger angle, γ.
Figure 7.20 Typical meshing of the tip gap (TG) generated by TurboGrid.
188
Chapter 7
The pressure contours at the tip and the leading edge of the turbine blade with
different values of tc staggered at 45°, 55° and 70° are shown in Figures A.5 to A.7 (see
Appendix-7). Variations of pressure from inlet to outlet with blades staggered at 45°, 55º
and 70º and different values of tc (TGs) are shown in Figures A.8 to A.10 (see Appendix
A-7).
Based on pressure and torque data deduced from 3D CFD simulations of a full scale
Denniss-Auld turbine, the effect of tc on turbine performance was evaluated in terms of η,
CT and P* ( see, Figures 7.21 to 7.23).
0.8
γ = 45º
γ = 55º
0.6
γ = 70º
η
0.4 0%
1%
0.2 2%
3%
0mm
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
φ
Figure 7.21 Effect of tip clearance, tc, on efficiency, η, of the full scale Denniss-Auld
turbine model with blades staggered at 45°, 55° and 70°.
Thus, it is seen from Figure 7.21 that, as expected, the maximum values of turbine
efficiency, η, were achieved by turbine models without tip clearance for all stagger
angles. Moreover, the efficiency was more noticeably affected by the magnitude of tc at
larger values of γ (55º and 70º). The efficiency of the turbine model with γ = 45º was
decreased approximately by 2%, 2% and 3% with increase of tc by 1%, 2% and 3%
respectively. The largest decrease of η was attained by the turbine model with γ = 70º,
where increase of tc by 1%, 2% and 3% led to reduction of efficiency approximately by
3%, 5% and 6%, respectively.
189
Chapter 7
1.2 γ = 45º
0.9
γ = 55º
CT
0.6 0%
1%
γ = 70º
0.3 2%
3%
0mm
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
φ
Figure 7.22 Effect of tip clearance, tc, on torque coefficient, CT, of the full scale Denniss-
Auld turbine model with blades staggered at 45°, 55° and 70°.
3.5 γ = 45º
2.8
2.1 γ = 55º
P* 0%
1.4 1%
γ = 70º
2%
0.7 3%
0mm
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
φ
Figure 7.23 Effect of tip clearance, tc, on pressure coefficient, P*, of the full scale
Denniss-Auld turbine model with blades staggered at 45°, 55° and 70°.
Analysis of torque, CT, and pressure, P*, coefficients, revealed the same trend, these
parameters were decreased with increase of tc for all stagger angles. The largest reduction
190
Chapter 7
of CT was demonstrated by the turbine model with the largest stagger angle of 70°and
largest tc =3%, the torque coefficient was decreased approximately by 15% compared to
the model without tc (Figure 7.22). For turbine models with the same γ, but with smaller
values of tc (1% and 2%), CT was reduced approximately by 7% and 11% respectively. In
contrast to CT, the largest reduction of P* was shown by the model with the smallest
stagger angle of 45°, the pressure coefficient was decreased approximately by 2%, 5%
and 8% with increase of tc by 1%, 2% and 3%, respectively (Figure 7.23).
φ = 2.16
tc= 1% tc= 3%
Figure 7.24 Surface streamlines over the suction side of blades with tc =1% and tc = 3%
staggered at 45°. The flow coefficient, φ, was equal to 2.16.
In addition to results presented above, the present author tested the effect of tc on flow
separation on the suction side of the staggered blades, which can advance the onset of
stall of the full scale Denniss-Auld turbine. According to Torresi et al., (2008), a CFD
model of the Well turbine with fixed blades and TG=1% demonstrated deep flow
separation over a large portion of the suction surface of the blades at different angles of
attack. In contrast to the Wells turbine model with TG=1%, the authors found that larger
values of tip clearance, such as 5% and 10% did not induce a flow separation on the
suction side of the turbine blades. The most suitable configuration of the Denniss-Auld
turbine for the flow separation test was a model with the smallest stagger angle of 45°.
The test was conducted at the highest value of φ = 2.16. Surface streamlines over the
suction side of the blades with tc = 1% and tc = 3% staggered at 45° are shown in Figure
7.24 where it is seen that none of the models demonstrated significant flow separation on
191
Chapter 7
the suction side of the blades. A similar trend was shown by the models with large values
of γ.
7.3.2.2 Hub to tip ratio, h

In order to study the effect of the hub to tip ratio, h, on turbine performance, the
following values of h: 0.51, 0.62 and 0.9 were tested. The actual turbine had a hub to tip
ratio of 0.77, which was taken as a baseline (bl) in the present study. The hub to tip
values of 0.51, 0.62 and 0.9 were achieved by varying the radius of the hub and keeping
the radius of shroud (786mm) unchanged. Variations of h were modelled at stagger
angles of 45º and 70º in identical operating conditions typical for these angles. The
angular velocity of the rotational domain at each run was kept constant (36.6rad/s or
350RPM) and only the axial velocity at the inlet was varied. Tip clearance, tc, for all
simulations was set as 4.6mm.
Numerical results in terms of η, CT and P* are presented in Figures 7.25 to 7.27. It can
be seen that the highest efficiencies were demonstrated by models having h = 0.62, γ =
45º and h = 0.51, γ = 70º with peak value of 0.7 and 0.54 respectively (Figure 7.25). It is
also seen that the largest h = 0.9 provided the lowest η for both stagger angles.
Comparison results against the data based on baseline configuration (h = 0.77) revealed
that efficiency was increased approximately by 1% and 5% by the models with h = 0.62,
γ = 45º and h = 0.51, γ = 70º, respectively. The trend of increasing η with decreasing of h
is evident for the turbine model with the largest γ = 70º. However, the difference between
efficiency data based on turbine models with h = 0.51 and h =0.62 was less than 1%. On
the other hand, the model with γ = 45º did not follow this trend. The turbine efficiency
was decreased approximately by 4% by the turbine model based on the smallest value of
h = 0.51 compared to the baseline configuration. The analysis of η indicates that h = 0.62
was best for both stagger angles.
CFD analysis of the torque coefficients, CT, based on models with γ = 45º showed that
the configurations with h=0.77 (baseline) and h=0.62 performed better than
configurations with h=0.51 and h=0.9 (Figure 7.26). Comparison of data in the interval of
0.9 ≤ φ ≤ 1.57 revealed insignificant difference (less 1%) between results for h = 0.62 and
h = 0.77. However, at higher values of φ the difference become more noticeable reaching
approximately 5%. Analysis also showed that CT data based on the model with h = 0.51
were lower approximately by 16% compared to the configuration with h = 0.62. It is
192
Chapter 7
interestingly to note that the CT data produced by the model with h = 0.9 in the range of
0.91 ≤ φ ≤ 1.75 were lower approximately by 12% than data based on configuration with
h = 0.51. However, at higher values of φ the CT data shown by the model with h = 0.9
were higher approximately by 2% compared to configuration with h = 0.51. The CT data
based on models with γ = 70º demonstrated the same trend as models with γ = 45º.
0.8
γ = 45º
0.6 γ = 70º
η
0.4
h=0.51
h=0.62
0.2
h=0.77-bl
h=0.9
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
φ
Figure 7.25 Effect of hub to tip ratio, h, on efficiency of the full scale Denniss-Auld
turbine model with blades staggered at 45º and 70º.
3D CFD analysis revealed that the turbine model with γ = 45º provided the highest
pressure coefficients, P*, when the smallest value of h was tested (Figure 7.27). For
instance, the P* coefficients deduced from the model with h = 0.51 were approximately
by 10% higher than it was shown by the baseline configuration. It is also seen from
Figure 7.27 that increase of h in configuration with γ = 45º led to decrease of the P* data.
Moreover, the difference between the P* data based on the model with γ = 45º increased
with increase of φ. In contrast to the model with γ = 45º, turbine configuration with γ =
70º demonstrated the opposite behaviour of P*. It can be seen that decrease of h
decreased the values of P*. The P* data based on the model with the largest h = 0.9 were
higher approximately by 22% than the pressure coefficients based on the baseline
configuration (h = 0.77). On the other hand, the difference between P* data deduced from
the models with h = 0.62, h = 0.77 and h = 0.9 was less significant. It is also evident that
193
Chapter 7
distribution of P* data produced by models with γ = 70º was more linear than it was
demonstrated at a smaller stagger angle.
1.5
h=0.51
h=0.62 γ = 45º
h=0.77-bl
1.0
h=0.9
CT
0.5
γ = 70º
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
φ
Figure 7.26 Effect of hub to tip ratio, h, on torque characteristic of the full scale Denniss-
Auld turbine model with blades staggered at 45º and 70º.
5
γ = 45º
h=0.51
4 h=0.62
h=0.77-bl
3 h=0.9
P
* h=0.51_7
2
γ = 70º
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
φ
Figure 7.27 Effect of hub to tip ratio, h, on pressure characteristic of the full scale
Denniss-Auld turbine model with blades staggered at 45º and 70º.
194
Chapter 7
In summary, 3D CFD analysis revealed that efficiency of the actual turbine can be
improved if the baseline value of h is decreased. For the Denniss-Auld turbine, h = 0.62
seems to be the most suitable value. This is very close to the value of h = 0.6, which was
recommended by Raghunathan (1995) for design of a Wells turbine.
Technically the adjustment of h can be achieved by two methods: a) increasing the
diameter of shroud (duct) and leaving the hub diameter unchanged; or b) decreasing the
hub diameter and keeping the dimensions of shroud (duct) unchanged. Both these
scenarios lead to an increase in the blade length. However, it should be noted that
increasing the radial dimensions of shroud (duct) are not always feasible due to technical
limitations imposed by the size of the OWC plant. On the other hand, variation of the hub
diameter is restricted by the diameter of the electrical generator in situations where the
generator is located within the nacelle.
7.3.2.3 Rotor solidity, σ

As mentioned earlier, the rotor solidity, σ, is one of the most important design
parameters of the Wells turbine. Amongst the first researchers who reported the effect of
solidity on performance of the Wells turbine were Gato and Falcao (1988). Figure 7.28
shows the original results of Gato and Falcao for the rotor solidities of 0.29, 0.44 and
0.59. The non-dimensional pressure coefficient curves are shown with closed symbols
and the efficiency curves are plotted with open symbols. It is clearly seen that the
efficiency of the Wells turbine tends to increase with decrease of σ in the interval of φ >
0.08. On the other hand, the higher rotor solidity the higher pressure coefficients of the
Wells turbine. Raghunathan (1995) also reported regarding the direct relation between the
efficiency of the Wells turbine and rotor solidity, according to his study significant
reduction of the turbine efficiency occurs when σ > 0.5.
In this sub-section, the effect of rotor solidity, σ, on performance of a variable pitch
turbine similar to Denniss-Auld design is discussed.
Variation of σ was modelled by decreasing the number of blades from 21 to 8. Blade
numbers of 8, 9, 11, 13, 17, and 21 (baseline) corresponded to solidity values of 0.42,
0.47, 0.58, 0.68, 0.9 and 1.11, respectively. The effect of σ was tested using turbine
models with blades staggered at 45º, 55º and 70º in the same operating conditions typical
for these angles.
The maximum number of blades tested in the present study was intentionally
restricted to 21 since a higher number of blades would cause very significant blockage in
195
Chapter 7
the plane of the turbine rotor, which in turn will decrease the efficiency, and more than 21
would not be able to be realised in practice due to the need to accommodate a practical
blade pitch mechanism. All simulations were conducted with a representative tip
clearance of 4.6mm.
Figure 7.28 Efficiency and pressure coefficient curves for Wells turbine published by
Gato and Falcao (1988). Symbols show experimental results and lines are from their
numerical analysis. The parameter U* represents the flow coefficient, φ.
Results of the 3D CFD simulations of a turbine with γ = 45º are presented in Figures
7.29 to 7.31. It can be seen that the baseline design (σ = 1.11) was more efficient than
models with lower values of σ in the interval of 0.9 ≤ φ < 1.18 (Figure 7.29). It is also
seen that at higher φ > 1.18 the efficiency of the baseline configuration significantly
decreased. At φ > 1.37 the baseline turbine model demonstrated the lowest η compared
to other models tested in study. In general, CFD analysis showed that decrease of σ up to
critical value increases the efficiency of the turbine with blades staggered at 45º in the
range of φ > 1.18. It is clearly seen that the critical value of σ was 0.58. Further decrease
of σ did not increase turbine efficiency. Comparison of CFD results in the interval of 0.98
≤ φ < 1.37 showed that the efficiency of the model with σ = 0.68 was approximately by
2% higher compared to the model with σ = 0.58. However, in the range of φ > 1.37, the
model with lower solidity provided efficiency by 1% higher than it was demonstrated by
the turbine configuration with higher solidity. On the other hand, in the interval of φ ≥
196
Chapter 7
1.18 the model with σ = 0.68 increased efficiency approximately by 3% compared to the
baseline design (σ = 1.11).
CFD analysis of torque coefficients, CT, (Figure 7.30) showed that the highest torque
characteristics were produced by the rotors with σ = 0.58 and 0.68. The lowest CT data
was demonstrated by the baseline configuration (σ = 1.11). The effect of σ on the torque
characteristics was especially noticeable at higher values of φ. However, at the lowest
values of φ the effect of σ on torque was minimal.
As expected, the decrease of σ decreased the pressure coefficients, P*, (Figure 7.31).
The model with the lowest solidity, σ = 0.42 provided the P* data, which were
approximately by 39% and 63% lower compared the models with σ = 0.68 and σ =
1.11(baseline), respectively.
Similar trend was observed for the model with the stagger angle of 55º. However, 3D
CFD modelling of turbines with blades staggered at 70° did not demonstrate any
advantages of low solidity rotors. The full results of the 3D CFD simulations of the
models with blades staggered at 55º and 70º can be seen in Appendix A-8.
0.8
σ = 1.11
σ = 0.9
γ = 45º
0.7 σ = 0.68
σ = 0.58
η σ = 0.47
0.6 σ = 0.42
0.5
0.4
0.0 0.5 1.0 1.5 2.0 2.5
φ
Figure 7.29 Effect of rotor solidity, σ, on efficiency, η, of the turbine model with blades
staggered at 45º.
197
Chapter 7
1.5
σ = 1.11
γ = 45º
σ = 0.9
σ = 0.68
1.0 σ = 0.58
σ = 0.47
CT σ = 0.42
0.5
0.0
0.0 0.5 1.0 1.5 2.0 2.5
φ
Figure 7.30 Effect of rotor solidity, σ, on torque characteristic, CT, of the turbine model
with blades staggered at 45º.
4
σ = 1.11 γ = 45º
σ = 0.9
3 σ = 0.68
σ = 0.58
*
P σ = 0.47
2 σ = 0.42
0
0.0 0.5 1.0 1.5 2.0 2.5
φ
Figure 7.31 Effect of rotor solidity, σ, on pressure coefficient, P*, of the turbine model
198
Chapter 7
In summary, CFD analysis of the rotors with different number of blades staggered at
45° and 55° indicated in favour of decreasing the rotor solidity up to 0.68 (or N = 13).
This approach should provide more efficient operation of the turbine through the range of
φ typical for stagger angles of 45° and 55°.
7.3.2.4 Blade profile

The effect of blade profile on turbine performance for the two types of blade (C and
D) shown in Figures 7.32 and 7.33 were tested. A full CFD optimisation of blade profile
was beyond the scope of the present project; nevertheless, it was seen to be important to
investigate the characteristics of these two types of blades which had previously been
implemented practically by Oceanlinx. The Blade C profile was the shape implemented
in the full scale Denniss-Auld turbine and Blade D was based on the shape of the blades
of the laboratory scale Denniss-Auld turbine (as described in Chapter 4). The maximum
thickness, τ, and the chord length, c, of both blade types were 15mm and 230mm,
respectively.
Figure 7.32 Type C profile.
Figure 7.33 Type D profile.
Performance of turbine models with blades of C and D types and tc = 4.6mm were
compared in identical operating conditions using the following stagger angles: 45º, 55º
and 70º. The results of the comparison are presented in Figures 7.34 to 7.35.
It can be seen that the use of C blades appears to shift the efficiency characteristic to
lower values of flow factor, φ, compared to the model based on blades of D type (Figure
7.34). Analysis showed that the models with blades of both types staggered at 45°
demonstrated the same value of peak efficiency equal to 0.69, but at different values of φ.
The peak efficiency of 0.69 was attained by the model based on the blades of C type at φ
= 1.06, while the model with blades of D type achieved this efficiency at φ = 1.18. It also
can be seen that at low values of φ from 0.9 to 1.06, the model with C blades provided
efficiency approximately by 5% higher than the model with D blades. On the other hand,
199
Chapter 7
at higher values of φ from 1.18 to 2.16, the efficiency demonstrated by the model with
blades of D type was approximately by 2% higher than it was shown by the model with C
blades. Analysis of η data of the turbine models with γ = 55° revealed that the model with
blades of both types had peaks of efficiency at the same value of φ = 1.57, but model with
C blades provided efficiency by 1% higher than the model with D blades. The efficiency
shown by the model based on C blades staggered at 55° at low values of φ ≤ 1.57 was
higher approximately by 8% than it was shown by the configuration with D blades. At
interval of higher values of φ from 1.96 to 2.75, the model with D blades was
approximately by 1% more efficient than the model with C blades.
The most significant effect of the blade shape on the turbine efficiency was shown by
both models with blades staggered at 70°. The model based on the blades of C type
provided efficiency approximately by 11% higher compared to the turbine configuration
with blades of D type.
0.8
C γ = 45º
D
γ = 55º
0.6 C_5
γ = 70º
η
0.4
0.2
0.0
0 1 2 3 4
φ
Figure 7.34 Effect of blade profile on efficiency, η, of Denniss-Auld turbine. Type C
(closed symbols/solid lines) and type D (closed symbols/broken lines).
200
Chapter 7
1.2
C γ = 45º
D
0.8
γ = 55º
CA
0.4
γ = 70º
0.0
0 1 2 3 4
φ
Figure 7.35 Effect of blade profile on input coefficient, CA, of Denniss-Auld turbine.
Type C (closed symbols/solid lines) and type D (closed symbols/broken lines).
Comparison of the turbine performance in terms of input coefficient, CA, and torque
coefficient, CT showed that the turbine rotor with C blades was significantly superior to
the model with D blades throughout the entire range of φ tested in the study (Figures 7.35
to 7.36). For instance, CA and CT coefficients based on the model with C blades staggered
at 45° were approximately by 24% and 25% higher, respectively than it was
demonstrated by the model with D blades. This was an interesting result as blade profile
C was seen at the time of the development of the full scale turbine as being something of
an “old style” of turbine blade while profile D had been developed and trialled more
recently by Auld and Finnigan (2003).
Effect of blade profile on turbine performance can be explained by analysing the
variations of angle of attack, α, at which the resultant velocity, WR, acted on the blades
shown in Figure 7.37. It can be seen that, at low values of φ, the blades of D type had
negative angles of attack, α, which were significantly large than it was shown by the C
blades. In turn, the large negative angles provoked the flow separation on the pressure
surface of D blades as shown in Figures 7.38 to 7.40. In contrast to D blades, the blades
of C type operated without flow separation. It is also seen that increase of φ eliminated
the flow separation on the pressure surface of D blades for all stagger angles.
201
Chapter 7
1.2 γ = 45º
C
D
C_5
γ = 55º
0.8
CT
0.4 γ = 70º
0.0
0 1 2 3 4
φ
Figure 7.36 Effect of blade profile on torque coefficient, CT, of Denniss-Auld turbine.
Type C (closed symbols/solid lines) and type D (closed symbols/broken lines).
1 γ = 45º γ = 55º
0
-1
-2 γ = 70º
-3
αο -4
-5
-6
-7 C
-8 D
-9 C_4
-10
-11
0 1 2 3 4
φ
Figure 7.37 Variations of α at which the resultant velocity, WR, acted on the blades.
202
Chapter 7
γ = 45º
Type C Type D
φ = 0.9
Type C Type D
φ = 1.18
Figure 7.38 Isometric views of the pressure surfaces of C and D blades staggered at 45º
at different flow coefficients, φ.
γ = 55º
Type C Type D
φ = 1.18
Type C Type D
φ = 1.57
203
Chapter 7
γ = 70°
Type C Type D
φ = 2.16
Type C Type D
φ = 2.94
The main reason of the flow separation on the pressure surfaces of the D blades
shown in Figures 7.38 to 7.40 can be the shape of leading edge. It is clearly seen from
Figure 7.33, that the radius of the leading edge of D blade is noticeably smaller compared
to the blade of C type shown in Figure 7.32. As a result of smaller radius at the leading
edge, the model with blades D had larger negative angles of attack and demonstrated the
flow separation at lower values of φ than the model with blades C designed with the
larger radius of leading edge. This consequently led to efficiency reduction at lower
values of φ shown by the model with the smaller radius of leading edge (type D). The
similar trend was reported by Setoguchi et al., (2003a), who investigated the effect of
blade profile on performance of the Wells turbine. It is also evident from Figure 7.37 that
the rotor with blades of C type was able to provide more smooth flow conditions through
the range of φ tested in study. Despite some efficiency reduction shown by the rotor with
C blades staggered at 45° and 55° at higher values of φ, the overall performance shown
by this turbine configuration indicates in favour of C blades rather than D blades.
204
Chapter 7
7.3.2.5 Blade thickness ratio, τ

As mentioned earlier, the blade thickness ratio, τ, is one of the key design parameters
of the conventional Wells turbine. The results, which were reported in several works
(Raghunathan, 1995, Setoguchi et al., 2003a, Takao et al., 2006, Thakker and Abdulhadi,
2007) indicated in favour of thicker symmetrical blades such as NACA0020, which
provided the highest turbine efficiency as in steady-state as well as sinusoidal flow
conditions (e.g. Figure 7.41). In light of this finding, it was interesting to investigate the
effect of blade thickness ratio, τ, on performance of the full scale Denniss-Auld turbine.
For this purpose, the baseline thickness ratio of blade C equal to 0.06 was decreased and
increased by 20%, which gave the values of 0.049 and 0.073 respectively. The profiles of
blades C with thickness ratios of 0.049, 0.06 (baseline) and 0.073 are shown in Figures
7.42 to 7.44. 3D CFD models of the rotors with cambered blades (type C), which had tc =
4.6mm and different thickness ratio, τ, were tested at stagger angles of 45º and 70º in
operating conditions typical for these angles.
Figure 7.41 Effect of blade profile on efficiency characteristics of the Wells turbine: (a)
steady-state, (b) sinusoidal flow conditions (Setoguchi et al., 2003a).
Figure 7.42 Type C profile: blade thickness ratio, τ = 0.049.
Figure 7.43 Type C profile: blade thickness ratio, τ = 0.06 (baseline).
205
Chapter 7
Figure 7.44 Type C profile: blade thickness ratio, τ = 0.073.
Results of 3D CFD modelling are shown in Figures 7.45 to 7.47. It can be seen that
the effect of increasing blade thickness was to decrease the peak efficiency achievable
and to shift the location of the peak to higher values of flow factor, φ. For example, for γ
= 45º decreasing τ by 20% from 0.06 to 0.049 led to an increase in peak efficiency of
72% while the location of the peak efficiency shifted from φ = 1.06 to φ =0.9
0.8
γ = 45°
0.049
0.6 0.06
γ = 70°
η
0.4
0.073 0.049
0.2 0.06
0.073
0.0
0 1 2 3 4
φ
Figure 7.45 Effect of τ on efficiency of turbine model with C blades staggered at 45° and
70°. Baseline curves (closed symbols/solid lines).
It is evident that in general for the Denniss-Auld turbine, thicker blades are less
efficient than thinner ones. This is contrary to the situation in the conventional Wells
turbine with fixed symmetrical blades (see, Figure 7.41) where thicker blades are
preferable.
206
Chapter 7
1.6
γ = 45° 0.049
1.2
0.06
CT
0.8 0.073
0.049
γ = 70°
0.06
0.4
0.073
0.0
0 1 2 3 4
φ
Figure 7.46 Effect of τ on CT of turbine model with C blades staggered at 45° and 70°.
Baseline curves (closed symbols/solid lines).
1.8
γ = 45° 0.049
1.2
0.06
CA
0.073
0.6
0.06
γ = 70°
0.049 0.073
0.0
0 1 2 3 4
φ
Figure 7.47 Effect of τ on CA of turbine model with C blades staggered at 45° and 70°.
Baseline curves (closed symbols/solid lines).
207
Chapter 7
7.3.3 Adjustment of turbine design parameters

The 3D CFD results presented in the previous section have shown that the efficiency
of a variable pitch turbine similar to Denniss-Auld turbine is affected more noticeably by
variation of the following design parameters: tip clearance, tc; hub-to-tip ratio, h, and
number of blades, Ν. It was found that the model of Denniss-Auld turbine with minimal
value of tc provides the highest efficiency for all stagger angles tested in study. It was
also discovered that the value of h = 0.62 is more favourable for a Denniss-Auld rotor
than the baseline value of 0.77, especially for larger stagger angles. 3D CFD analysis also
has demonstrated that decrease of rotor solidity, σ, up to certain value increases
efficiency of a variable pitch turbine similar to Denniss-Auld design. All these findings
are in line with researches by previous authors (e.g. Gato and Falcao (1988),
Raghunathan (1995), Torresi et al. (2007)).
3D CFD simulations of the rotors with blades of different profile (C and D) have
indicated in favour of employing the cambered blades of type C. Moreover, CFD analysis
has showed that the baseline blade thickness of 0.06 has provided optimum performance
of a variable pitch turbine with blades of type C staggered at angles of 45º and 70º.
Taking this information into account the geometry of the full scale Denniss-Auld turbine
configuration can be tuned to improve its overall performance.
In this section, the results of 3D CFD simulations of a turbine model with geometry
adjusted to increase the efficiency of a variable pitch turbine similar to Denniss-Auld
design are discussed. The turbine design parameters were adjusted as follows:
• Tip clearance, tc = 2.3mm (or 1% of the length of c);
• Hub-to-tip ratio, h = 0.62;
• Number blades, N = 13 (or σ = 0.75);
• Chord length, c: baseline value of 230mm;
• Blade thickness ratio: baseline value of 0.06.
The turbine model with this adjusted geometry was tested using the stagger angles:
45º, 55º and 70º in operating conditions typical for these angles. The average number of
structured elements for each computational domain consisting of 3 parts (S-1, R-1 and S-
2) was 1.3×106. The structured grids were generated by TurboGrid.
Following initial simulations on the PC the definition files were run on the cluster
computer. Comparison of CFD results based on a variable pitch turbine with adjusted
208
Chapter 7
configuration and original (baseline) design of the full scale Denniss-Auld turbine are
presented in Figures 7.48 to 7.50.
It is seen from Figure 7.48 that the tuned configuration noticeably increased turbine
efficiency compared to the baseline design through a relatively wide range of φ. For
instance, in the interval of 1.0 < φ ≤ 2.75 the turbine efficiency attained by the adjusted
rotor configuration with blades staggered at 45º and 55º was approximately 5.5% higher
than shown by the baseline design. In the interval of φ > 2.75, the adjusted model with
the largest stagger angle of 70º also increased efficiency on average by ~3% compared to
the original turbine.
0.8 Baseline
γ = 45°
γ = 55° Adjusted
0.6
γ = 70°
η
0.4
0.2
0.0
0 1 2 3 4
φ
Figure 7.48 A comparison between efficiencies, η, based on the original Denniss-Auld

turbine geometry and a turbine with the adjusted configuration.
209
Chapter 7
1.2 γ = 45°
Baseline
Adjusted
0.8
γ = 55°
CA
0.4
γ = 70°
0.0
0 1 2 3 4
φ
Figure 7.49 A comparison between input coefficients, CA, based on the original Denniss-
Auld turbine geometry and a turbine with the adjusted configuration.
1.5 Baseline
Adjusted γ = 45°
1.0 γ = 55°
CT
0.5
γ =
0.0
0 1 2 3 4
φ
Figure 7.50 A comparison between torque coefficients, CT, based on the original
Denniss-Auld turbine geometry and a turbine with the adjusted configuration.
210
Chapter 7
It is important to note that the starting characteristics (CA) demonstrated by the

original turbine were not lost with a decrease of σ in adjusted rotor, the torque
coefficients, CA, achieved by the tuned configuration with blades staggered 45º and 55º
exhibited similar magnitudes as a baseline model (Figure 7.49). Even the adjusted rotor
with the largest γ = 70º demonstrated same magnitude of CA coefficients as the baseline
turbine in the interval of φ > 2.75. Analysis of CT data also revealed superiority of the
adjusted rotor with blades staggered 45º and 55º in the interval of 1.16 < φ ≤ 2.75 (Figure
7.50). The torque characteristics, CT, in this interval of φ were increased by the adjusted
turbine approximately by 9% compared to the original turbine configuration. The
adjusted rotor with largest γ = 70º provided also higher torque characteristics in average
by 3% than shown by the baseline turbine in interval of φ > 2.91.
In summary, adjustment of design parameters such as: tip clearance, tc, hub-to-tip
ratio, h, and rotor solidity, σ, can noticeably improve the performance of the full scale
Denniss-Auld turbine through the relatively wide range of φ. In particular, the turbine
efficiency can be increased by the adjusted rotor with tc = 2.3mm, h =0.62 and σ = 0.75
(or N=13) approximately by 5.5% compared to the original design of the Denniss-Auld
turbine in the most efficient operational range of 1.0 < φ ≤ 2.75. Even at higher φ > 2.75,
the adjusted rotor with blades staggered at 70º can provide higher efficiency in average
by 3% than the baseline configuration.
7.3.4 Turbine cycle efficiency

For the purpose of completeness the issue of instantaneous turbine efficiency versus
overall cycle efficiency of an OWC turbine should be noted. The work presented in this
thesis is based throughout on turbine performance being assessed at a single flow rate.
However, in practice a turbine must be optimised with respect to the full range of
expected flows for a variety of sea conditions. This is a significant challenge to the
designer since the range of parameters in variable geometry turbines that can be
influenced is potentially very large. For example, controlling blade angle will directly
affect both pressure drop across the turbine rotor and efficiency over the course of the
wave cycle. Due to limitations in time available this complex issue was not addressed in
the present study, however turbine designers should be aware that it represents a key
challenge to the maximisation of overall OWC efficiency.
211
Chapter 7

The non-dimensional pressure coefficient, P*, deduced from 3D CFD modelling of
the Wells turbine model with variable pitch blades demonstrated good agreement with
experimental data published by Tease (2003). In addition, variations of torque coefficient,
CT, with variations of a pitch angle have been successfully modelled for a Wells turbine
with the same variable geometry as reported by Tease (2003). A comparison of predicted
performance characteristics of the Wells turbine with variable pitch blades deduced from
BEM and 3D CFD simulations showed good agreement in the pre-stall region.
Performance data deduced from 3D CFD modelling of a 1/8th sector of the actual
turbine by the present author have been compared with the experimental results reported
by Finnigan and Auld (2003) for their small scale Denniss-Auld turbine model. Turbine
efficiency at small stagger angles was predicted relatively accurately, but high blade
stagger angles gave less favourable agreement. 3D CFD analysis has revealed noticeable
discrepancies between the predicted and experimental torque coefficients, CT, for all
stagger angles tested in study. The main cause of this may be due to the fact that the 3D
CFD models were run without inclusion of frictional and/or other losses produced by the
apparatus employed to measure the torque during the tests in wind tunnel. The difference
between the predicted and measured input coefficient, CA, at a stagger angle of 20° can be
attributed to sensitivity of turbine blades with smaller stagger angles to variation in tip
clearance, tc. 3D CFD analysis of the full turbine with blades staggered at 20° coupled to
the nacelle has confirmed this suggestion as well as the magnitude of other performance
characteristics predicted by the model based on 1/8th part of the actual turbine.
One of the interesting results obtained from 3D CFD analysis of the Denniss-Auld
turbine is determination of the optimal rotor solidity, σ. Simulations of the rotors with
different number of blades staggered at 45° and 55° demonstrated that a rotor solidity of
~0.68 (or N = 13) provides the most efficient operation of the variable pitch turbine
through the range of φ typical for these stagger angles. It was also found that further
decrease of σ in rotors with blades staggered at 45° and 55° did not improve the turbine
efficiency. However, when turbine blades staggered at 70° the highest efficiency was
shown by the rotor with the baseline number of blades (N = 21) or rotor solidity of 1.11.
Since the sea-trial tests showed that the most efficient operation of the Denniss-Auld
turbine was attained with blades staggered from 45° to 55°, it indicates that solidity
adjustment should be focused on this range of γ.
212
Chapter 7
Based on the results presented in sections 7.3.2 and 7.3.3 the following important
conclusions regarding potential improvements to efficiency can be made:
• Efficiency improvements, on the order of 5.5% are possible by the present limited
steps taken to optimise the current design of the Denniss-Auld turbine.
• A small tip clearance, tc, (~1% of the length of chord) is the most preferable for
efficiency improvement.
• Adjustment of the hub-to-tip ratio, h, of the actual turbine to the value of 0.62 can
improve performance of the Denniss-Auld turbine.
• Fewer blades (e.g. ~13) are preferred to give lower rotor solidity, σ, and
consequently increased the turbine efficiency. A rotor solidity of 0.75 based on N
= 13 and h = 0.62 has shown the most efficient performance of the variable pitch
turbine similar to the Denniss-Auld design.
• It would appear that the cambered blades of type C are preferable to the flat
blades of type D, at least in the full scale Denniss-Auld turbine.
• Thicker blades of type C shift the turbine efficiency to higher values of φ, while
thinner blades to the interval of lower flow coefficients, φ.
213
Chapter 8
CHAPTER 8 - CFD ANALYSIS OF OWC PNEUMATIC SYSTEM

The three-dimensional CFD results presented in the previous chapter confirmed the
utility of the ANSYS CFX-11 code in modelling the performance of axial flow turbines
in isolation from the rest of the OWC system. However, an actual OWC turbine is always
coupled to other OWC components such as the OWC chamber, nozzle-diffuser, electrical
generator and grid network. The size, configuration and control of these components all
play an important part in determining the performance of the OWC turbine. In this
chapter, results from CFD simulations of a full scale 3D model of a Denniss-Auld turbine
coupled with a complete pneumatic system are presented and discussed. The system
studied was similar in configuration and dimensions to the Oceanlinx OWC plant tested
at Port Kembla (Figure 8.1). Additionally, the internal flow field in the nozzle-diffuser
(air duct) was analysed and the implications for OWC system design are discussed,
particularly in regard to the need for uniform flow at the inlet to the turbine rotor.
Figure 8.1 Oceanlinx OWC plant tested at Port Kembla, Australia.
8.1 Performance analysis of a complete OWC pneumatic system

One of the objectives of the present study was to analyse the performance of a
complete OWC pneumatic system with aid of 3D CFD modelling. In order to achieve this
task, firstly, a turbine model with 21 blades having the dimensions of a full scale turbine
214
Chapter 8
was built (see, Figure 8.2). The mesh was developed such that the overall number of
unstructured elements was 2.6 × 106. Secondly, the OWC chamber was linked with the
converging and diverging sections of a nozzle-diffuser and a nacelle created using
DesignModeller™. The dimensions of these components were based on the dimensions
of Oceanlinx OWC plant tested at Port Kembla. All OWC components were also meshed
by unstructured elements with a total number of 3.2×106. The third step was carried out
using the pre-processor of ANSYS CFX, where the OWC components and a turbine were
coupled to each other by general grid interfaces with a Frozen Rotor frame change. A
rotational reference frame with an angular velocity equal to 36.6rad/s (or 350RPM) was
specified for the parts of the computational domain comprising the turbine rotor, hub, and
shroud. The rest of the computational domain was modelled using a stationary reference
frame. Non-slip smooth walls were used for the blades, nacelle, hub and shroud of the
turbine and also on the walls of the nozzle-diffuser and OWC chamber. In addition, the
boundaries of the shroud were set to rotate in the opposite direction to the blades at the
appropriate velocity. In order to simulate the exhaling-inhaling flow cycles, the mass
flow rate was specified at the inlet, and the static pressure at the outlet was set to one
atmosphere. The simulations were run with various values of mass flow rates to provide
the necessary variation of the axial flow velocity through the rotor. However, due to the
prohibitively large CFD runs times required to complete a full transient analysis of the
system over a full wave cycle, a series of steady-state simulations were carried out at
discrete intervals over the expected range of axial flow velocities. The overall
performance of the complete OWC pneumatic system was tested with the turbine having
the blades staggered at 45º. All results presented in this chapter were obtained using the
cluster computer. It should be noted that supporting legs for turbine, electrical generator
and nacelle were not included in the geometry shown in Figure 8.2 and consequently
CFD results presented in this and following sections do not take into account their
influence. This was so as to limit the complexity of the computational domains used in
the present study so to ensure practicable computer run times. However, the presence of
supporting legs can influence turbine performance and in this respect, this issue may be
considered as a potential topic of future research.
215
Chapter 8
OWC chamber
Nozzle (converging section)
Rotor
Diffuser (diverging section)
Nacelle
Inlet: Mass flow rate (kg/s)
Outlet: Pst = 1atm
Bellmouth
Figure 8.2 Isometric view of a full scale Denniss-Auld turbine with blades staggered at
45º incorporated in OWC wave energy converter (exhalation cycle).
CFD analysis of the complete OWC pneumatic system shown in Figure 8.2 was
carried out assuming that volume flow rates in both exhaling and inhaling cycles were
identical. Air density, ρa, was set at 1.20kg/m3. Mass flow rates of 23.7, 33.5, 42.7 and
52.5kg/s were set as the boundary conditions at the inlet and the outlet so as to provide
axial velocities, VA, through the rotor of 25, 35, 45 and 55m/s. Typical results of pressure
variation through the length of the pneumatic system are presented in Figures 8.3 and 8.4.
Analysis of the pressure distribution in the pneumatic system of the OWC showed that
the system demonstrated very similar pneumatic losses in the exhaling and inhaling
cycles as shown in Figure 8.5.The pressure losses across the 90º bend from OWC
chamber to the air duct during the exhaling cycle for all mass flow rates were very small
(e.g. ~ 4Pa for the mass flow rate of 33.5kg/s where overall pressure loss was 1.83kPa). It
is evident that the total pressure losses due to the presence of the bend can be ignored in
terms of pneumatic losses. However, kinetic energy losses from the system through air
exiting the nozzle/diffuser can clearly be significant.
216
Chapter 8
18
9 10
2 3 4 5 6 7 8 11 12 13 14 15 16 17
Figure 8.3 Typical pressure contour during exhalation cycle. The mass flow rate through
the rotor was 33.5kg/s and air density was 1.20kg/m3.
Pst (+)
2 18
9 10
Ptot (+) 3 4 5 6 7 8 11 12 13 14 15 16 17
Pst (-)
Ptot (-) 1
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
102500
102000
101500 Patm
(Pa) 101000
100500
100000
99500
0 4 8 12 16 20
Location (m)
Figure 8.4 Typical pressure gradient diagram of the OWC during exhaling (+) and
inhaling (-) cycles. The mass flow rate through the rotor was 33.5kg/s.
217
Chapter 8
0.002%
10%
10%
41% 41%
49% 49%
Bend
Nozzle-nacelle Nozzle-nacelle
Rotor Rotor
KE losses KE losses
a) b)
Figure 8.5 Relative total pressure loss in OWC wave energy converter during a) exhaling
and b) inhaling (mass flow rate = 33.5kg/s).
Figure 8.6 shows where energy is lost/converted in the nozzle/rotor/diffuser system

based on the 3D CFD analysis. It can be seen that at lower volume flow rates, Q, up to
approximately 30m3/s at least half of the available pneumatic power was lost as a result
of friction upstream of the rotor. The proportion of the frictional losses due to the nozzle-
nacelle significantly decreased at higher axial flow velocities.
300
Exhaling
Ppneu_tot
Inhaling
200
Ppneu (kW)
Rotor
100 Nozzle/nacelle
KE losses
Bend
0
0 10 20 30 40 50
3
Q (m /s)
Figure 8.6 Pneumatic energy conversion/loss in a typical OWC wave energy converter
(similar to the Oceanlinx Plant).
218
Chapter 8
8.2 Analysis of the internal flow field in a nozzle-diffuser system

As shown in Figure 8.2, the Oceanlinx OWC plant employed a nozzle-diffuser
arranged horizontally i.e. with axis at 90° to the main-flow in the OWC chamber. This
duct arrangement facilitates practical access to the turbine mechanisms and electrical
generator for maintenance and repair, however, the 90° bend can cause non-uniform flow
in the diffuser. In turn, this non-uniformity can lead to asymmetry in the flow field
approaching the turbine rotor, in contrast to a vertical duct system. However, a vertical
duct system is not recommended since it causes several unattractive consequences
including a taller more visible structure; poor access for maintenance; and additional
thrust load due to rotor weight.
In this section, the airflow field in the horizontal duct system of the Oceanlinx OWC
plant is analysed and potential methods that can improve the velocity distribution, in
terms of uniformity in vicinity of the rotor are modelled and discussed.
8.2.1 Flow in horizontal air duct

Some of OWC plants such as the LIMPET (Scotland) and the Pico Plant in the
Azores (Portugal) introduced in Chapter 2 employed horizontal air ducts of constant
diameter (see for example Figure 8.7). It is interesting to compare the velocity
distribution in such duct arrangements with that in nozzle/diffuser systems. Figures 8.8 to
8.9 show velocity contours in the horizontal air ducts with the constant arbitrary chosen
diameters of 1.64m and 2.46m, respectively. It can be seen that the 90° turn from the
main flow produced a non-uniform flow field approaching the nacelle. The worst velocity
distribution in terms of non-uniformity was demonstrated by the duct with the larger
diameter. In contrast to the horizontal duct with constant diameter, the converging nozzle
section with an included angle of 7° produced a virtually uniform flow field throughout
the annulus (Figure 8.10). It is evident that use of a converging nozzle section improves
the airflow field approaching the turbine rotor. The similar principle is applied to achieve
a uniform velocity stream in the working section of wind tunnel (Wallis, 1961).
219
Chapter 8
Figure 8.7 Arrangement of Turbo Generation Equipment (LIMPET, 2002)
D = 1.64m
Location at 0.5m upstream
Vin = 0.46m/s
1.2m in diameter.
220
Chapter 8
D = 2.46m
Vin = 1.5m/s
1.2m in diameter.
D = 1.57m
7°
Vin = 1.02m/s
Figure 8.10 Velocity flow field in converging nozzle section with included angle of 7°.
Diameter at the outlet is 1.57m. Nacelle diameter is 1.2m.
221
Chapter 8
8.2.2 Flow in nozzle-diffuser in absence of the nacelle

Prior to consideration of the possible impact of the internal bodies on airflow in the
horizontal nozzle-diffuser, it was essential to get a clear picture of the velocity
distribution in absence of internal blockages.
9°
Vin = 2.5m/s
Figure 8.11 Velocity flow field in the horizontal nozzle-diffuser comprising the
converging and diverging sections with included angle of 9°. Velocity at the inlet was
2.5m/s.
The velocity contour taken on the XY plane (Z=0) showing the variation of the flow
field in the nozzle-diffuser in absence of a nacelle is presented in Figure 8.11. It can be
seen that the converging nozzle section produced a fairly uniform flow field at the turbine
location. On the other hand, significant flow separation as a result of a relatively large
included angle (9°) is clearly seen on the lower side of the diverging diffuser section.
This result confirms the findings published by Wallis (1961), who claimed that the flow
separation in a diffuser is usually confined to one side. Figure 8.12 gives an isometric
view of OWC converter and illustrates the air flow through the nozzle-diffuser using
streamlines.
222
Chapter 8
Flow separation
Figure 8.12 Illustration of the flow separation from the lower side of the diverging part
of the air duct. Complimenting results shown in Figure 8.20.
8.2.3 The length of nozzle-diffuser section

In order to decrease the overall costs of an OWC system it is important that a compact
nozzle-diffuser be used. To investigate the effect of nozzle-diffuser length on pneumatic
efficiency of OWC system the baseline length of the nozzle-diffuser section was
decreased by 2m. The length of the nacelle was left unchanged. The results of this study
are presented in Figure 8.13. It can be seen that the shorter nozzle-diffuser sections (5m)
provided the better pneumatic efficiency, ηpneu compared to the baseline length. The
pneumatic efficiency was increased approximately by 1.5% over entire range of Q tested
in study. The higher efficiency can be explained by the fact that the KE losses in shorter
nozzle-diffuser were approximately by 2% lower than in the baseline configuration. This
finding makes it possible to optimise the size of the actual nozzle-diffuser by reducing its
length at least by 4m.
223
Chapter 8
0.6
5m
7m
0.5
η pneu
0.4
0.3
0 10 20 30 40 50
3
Q (m /s)
Figure 8.13 Effect of nozzle-diffuser length on pneumatic efficiency, ηpneu of OWC
system.
8.2.4 Nacelle shape
The velocity distribution through the nozzle-diffuser with the baseline nacelle shape
is presented in Figure 8.14. It can be seen that the baseline nacelle shape produced non-
uniform flow field at the narrowest part of the duct. In contrast to the baseline shape, the
nacelle geometry with conical noses depicted in Figure 8.15 produced a significantly
more uniform flow field. This finding strongly indicates in favour of employing the
nacelle with conical noses in future projects. However, the optimum value of the angle
forming the conical shape must be determined in further research.
224
Chapter 8
No rotor
Vin = 1.02m/s
Figure 8.14 Velocity flow field in horizontal nozzle-diffuser with the baseline nacelle
shape (no rotor is presented). Velocity at the inlet to base of the OWC chamber was
1.02m/s.
No rotor
Vin = 1.02m/s
Figure 8.15 Velocity flow field in horizontal nozzle-diffuser with the nacelle having
conical noses (no rotor is presented). Velocity at the inlet to base of the OWC chamber
was 1.02m/s.
8.2.5 Difference between inhaling and exhaling

During inhalation cycle should be little flow distortion and a relatively uniform axial
velocity distribution at the turbine inlet was to be expected.
225
Chapter 8
Figures 8.16 shows the axial velocity contours on the XY plane (Z=0) and at the
annulus 0.5m upstream the rotor obtained at a mass flow rate of 23.7kg/s and turbine
blades staggered at 45º. It is clearly seen that the axial velocity distribution upstream to
the rotor was uniform. The same result was found at other mass flow rates tested.
Pst = 1atm
Location 0.5m upstream
Mass flow rate = 23. 7 kg/s X
Figure 8.16 Axial velocity field in a nozzle-diffuser during inhaling flow cycle. The
blades were modelled as rotating at 350RPM with stagger angle, γ = 45°.
The case of exhalation is presented in Figure 8.17. It is clearly seen that the
90°change of the airflow direction from the OWC chamber to the nozzle-diffuser did not
distort the axial velocity field approaching to the rotor. Analysis also revealed that the
axial velocity in all locations across the annulus was uniform. The same trend was shown
by the model of OWC system at other mass flow rates tested in study.
226
Chapter 8
Pst= 1atm
V = 0.46m/s
Massinflow rate =23. 7kg/s
Figure 8.17 Axial velocity field in nozzle-diffuser during exhaling flow cycle. The
blades were modelled as rotating at 350RPM with stagger angle, γ = 45°.

It was found that the reduction in pneumatic power available to the rotor of the OWC
system due to the 90° bend between the chamber and the horizontal air duct was
negligible, less than 0.01%.
3D CFD analysis revealed that the pneumatic efficiency of the OWC system based on
the dimensions of an actual OWC wave energy plant to a large extent depends on the
volume flow rate through the horizontal air duct. It was determined that about half of the
available pneumatic power is lost in the OWC system as a result of friction in the nozzle-
diffuser sections when the volume flow rate, Q, was less than ~30m3/s.
Visualisation of the velocity field in horizontal air ducts of constant diameter
confirmed the considerable impact caused by the turn of 90º on the circumferential
uniformity of the internal flow field. The flow field in the duct with the largest constant
diameter was distorted by the change of the flow direction the most. This is in contrast to
conical ducts with an included angle of 7º and a nacelle, which produced a virtually
uniform flow field.
The 3D CFD analysis demonstrated that the shorter nozzle-diffuser section (5m)
provided better pneumatic efficiency, ηpneu compared to the baseline length of 7m. The
pneumatic efficiency of the OWC system was increased by approximately 1.5% over the
227
Chapter 8
entire range of flow rates tested in the study. Furthermore, the nacelle with conical noses
produced a more uniform velocity distribution upstream of the turbine than the baseline
nacelle with rounded ends.
The 3D CFD simulations of both exhaling and inhaling airflows through the
horizontal nozzle-diffuser and turbine with rotating blades staggered at 45° showed that a
circumferentially uniform axial velocity distribution existed upstream of the rotor. In
other words the 90º change in airflow direction from the OWC chamber to the nozzle-
diffuser did not significantly distort the axial velocity field approaching the rotor.
228
Chapter 9
CHAPTER 9 - COMPARISON OF NUMERICAL ANALYSES AND

FIELD DATA FROM A FULL SCALE OWC PLANT
The important objective of the present research was to compare the numerical results
deduced from the BEM and CFD analyses with experimental data obtained during sea
trials of the entire OWC system at Port Kembla in 2006 by Oceanlinx (then known as
Energetech). This chapter presents the analysis of the experimental data as well as
comparison with results predicted by the numerical methods used in the present study. It
should be noted that the present author played no part in the design of the physical plant
or of the data acquisition system. The data was collected by the Oceanlinx team led by Dr
Ray Alcorn.
9.1 Analysis of field data from a full scale OWC plant

The results presented were for when the OWC plant tested by Oceanlinx at Port
Kembla, Australia, was set-up as a non grid connected, stand alone, operation. The full
scale Denniss-Auld turbine blades were pitch controlled using a servo drive/motor
combination attached to a linear actuator. The turbine was coupled to a 415V/250kW 6-
pole 25Hz 3 phase squirrel cage induction generator which was connected via a 440kW
rated variable speed regenerative drive to a controllable electric load. This arrangement
gave generative capability up to 500kW on peaks; however this was limited to 150kW
during the sea-trial tests due to onboard load restrictions.
In order to measure temperature, pressure, velocity, torque, wave height, electrical
power output, etc. the Oceanlinx OWC plant had an array of sensors (see, Figure 9.1). All
these sensors were connected to an Allen Bradley Contrologix PLC and were sampled at
a rate of approximately 10Hz. The pressure drop across the turbine was measured by a
differential pressure transmitter STX2100 from Druck Ltd shown in Figure 9.1 as
PIT0205. Other pressure sensors mounted in the nozzle-diffuser were LP1000 differential
gauges from Druck Ltd, which are shown in Figure 9.1 as PIT0101 to PIT0104 and
PIT0106 to PIT0109. It should be noted that both types of pressure sensors had adjustable
“damping”, from 0 to 38.4 seconds and from 0.01 to 2 seconds for STX2100 and
LP1000, respectively, which with could lead to some differences in the response time.
However, despite various enquiries it is not known to the present author what these
damping settings were in practice.
229
Chapter 9
The torque produced by the turbine was measured through the electronic generator
control system. The measured values of torque were used to calculate the power output of
the OWC system and efficiency of the full scale Denniss-Auld turbine.
The raw data on the turbine performance, air pressures and flow rates presented in this
document were generously provided by Oceanlinx and subsequently processed by the
present author and Prof. Paul Cooper. It should be noted that for reasons of commercial
confidentiality only a limited range of technical details of the OWC configuration and
experimental data can be released.
Figure 9.1 Sensor layout (provided by Oceanlinx).
The most convenient way of comparing data for a wide range of operating conditions
was to use non-dimensional parameters for efficiency, torque and pressure drop through
the turbine as a function of non-dimensional axial velocity or flow coefficient, φ. These
parameters were evaluated using the definitions (3.10, 3.11, 3.12. 3.13 and 3.14)
described in Chapter 3.
One of the critical issues regarding the performance evaluation of the full scale
Denniss-Auld turbine was determination of a true value of the mean axial air flow
velocity through the turbine. The analysis of the raw data revealed that there were
considerable inconsistencies between the various options for calculating air velocity.
230
Chapter 9
Considering various sensors that were available on the plant, the following potential
methods were available for determining volumetric air flow rate as a function of time:
a) From rate of change of the free surface sea level in the OWC chamber using
either: i) a sub-surface pressure sensor; or ii) an ultrasonic water level sensor;
b) Using a hot-film anemometer within the turbine ductwork;
c) Using static pressure data from various cross sections of the nozzle-diffuser
upstream and downstream of the turbine.
From analysis of the volume flowrate readings it was clear that use of the ultrasonic
and underwater pressure sensor traces for determining chamber free surface elevation was
not a viable option for determining volume flow rates of air through the systems. Even
when a sixth-order smoothing algorithm known as the Douglas-Avakian method
(Whitaker and Pigford, 1960) was used (as shown with the data presented in Figure 9.2),
the noise on the signals was too high to give a mean chamber free surface velocity – most
likely due to waves and turbulence on the free surface inside the chamber, and
interference due to spray and moisture in the chamber.
After detailed analysis of the raw data it was determined that the most reliable
velocity calculation method, for the sensor configuration available, was that based on
static pressure measurement at adjacent stations in the converging part of the horizontal
air duct (converging with respect to direction of flow) as if this was a venturi. The
following equation results from the Bernoulli equation (assuming no frictional losses
between the two pressure stations):
2∆p
V2 = (9.1)
  A 2 
ρ 1 −  2  
  A1  
 
After comparing all the different options for measuring air velocity (see Figure 9.2,
for example) it was suggested that the safest option is to use equation (9.1) with pressure
readings from sensors PIT0203/0204 when exhaling and PIT0208/0207 when inhaling.
Note: “inhaling” is taken to be the situation where the mean free surface elevation in the
OWC is decreasing, and “exhaling” is the converse situation.
Thus despite a great deal of effort undertaken to determine the axial velocity at the
turbine location, there is a deal of uncertainty as its accuracy. This should be born in
mind when revising the comparisons with the numerical data (presented below) and is a
231
Chapter 9
question that should be carefully addressed in assessment of all full-scale OWC

installations.
80
60
40
20 Flow 2 to 3
Q (m3/s)
0 Flow 3 to 4
-20 Flow 7 to 6
-40 Flow 8 to 7
Flow 9 to 8
-60
Anemometer
-80
326 328 330 332 334 336
Time (s)
Figure 9.2 An example of comparison of volume flowrates, Q, calculated using data
from various pressure sensors and hot-film anemometer in the full scale Oceanlinx OWC
Port Kembla plant. The numbers from 2 to 9 represent the pressure sensors.
Another very important aspect of turbine performance, which should be also born in mind
conducting analysis, was that turbine control system maintained a nominally constant
RPM. In this respect any uncertainty in turbine torque due to inertial effects during
operation were assumed not to be signifant. The control system in maintaining this
constant speed therefore meant that electrical power was either supplied to or generated
by the generator depending on the aerodynamic conditions at a given instant in time (see
Figure 9.3). The data presented below are plotted with a power threshold of 10kW. It is
clearly seen that there is a number of time intervals (e.g. from 305s to 309s) where power
output was negative. It is also interesting to note that the pressure drop across the rotor
during inhalation (negative ∆P) was higher compared to the exhaling cycle (e.g. in the
interval of time presented below approximately by 50%). In addition to that the volume
flow rate, Q, during inhalation was also higher than recorded in the exhaling cycle (e.g. in
the interval of time presented below approximately by 14%). It is evident that
combination of higher values of ∆P and Q provided higher pneumatic power available for
the turbine during inhaling flow cycles.
232
Chapter 9
One further issue which can add a degree of uncertainty in the evaluation of the full
scale practical turbine performance, and in particular its efficiency, comes from the
possible variation in air density during exhaling versus inhaling cycles. This variation
may arise, for example, due to differences in humidity between the air in the OWC
chamber and the ambient and because of the fact that the absolute pressure in the
chamber can be significantly above and below ambient for large wave heights. In reality
this variation would not be negligible, however, research on this issue was beyond the
scope of the present study and may be considered as an issue for future research.
AC Power Power calc'd from Torque/RPM

air flow rate m3/s pneumatic power
pressure across turbine efficiency
160 1.0
140
0.5
120
100 0.0
∆P (kPa), η
Power (kW)
80
-0.5
60
40 -1.0
20
-1.5
0
305 310 315 320 325 330
-20 -2.0
Time (s)
Figure 9.3 An example of comparison of different turbine characteristics as a function of

time (the power threshold is 10kW, i.e. efficiency only calculated for instances where this
power threshold was exceeded).
The experimental data presented below was gathered during the very first phase of
the testing of the Oceanlinx Port Kembla plant. The control algorithm for the blade
stagger angle activation mechanism was a simple square wave pitch variation with
respect to time. That is, for the given stagger angle, γ, under test, the control system set
the stagger angle to this constant value throughout the “exhaling” part of each wave
cycle. The blade stagger angle was then quickly reversed at the point at which air flow in
the turbine reversed, so as to provide the same (through negative) stagger angle during
the inhalation phase of the cycle. Thus, this relatively crude initial control algorithm
provided a constant stagger angle with which the numerical BEM and CFD results of the
233
Chapter 9
present study could be compared. Clearly the actual flow conditions within the turbine
and OWC system as a whole were far from steady state. This fact should be noted
throughout the following section.
Before the full set of experimental data is presented it is useful to consider just one
stagger angle (arbitrarily chosen here as γ = +/-45° for the purposes of illustration).
Figures 9.4 to 9.6 show the processed experimental data for the actual turbine with blades
staggered at +/-45° and with a power threshold of 10kW in terms of efficiency, η,
pressure coefficient, P*, and torque coefficient, CT. Use of a threshold value of 10kW
removes data recorded when the power output from the generator was small or negative
and various issues such as electrical and mechanical noise then rendered the efficiency
results unreliable. The length of data record was 660s. The experimental data presented
include data for both flow cycles: exhaling and inhaling (i.e. where inhaling indicates that
the air flow through the turbine is from the shore-side ambient to the OWC chamber). A
negative blade angle represents the situation during the inhalation part of a wave cycle.
The data were plotted separately for each wave cycle and arrows show the recorded
sequence of the data with respect to increasing time (see Figures 9.7 to 9.9). It can be
seen that all instantaneous experimental data for γ = +/-45° had significant scatter even
with the power threshold of 10kW (although use of the threshold significantly reduced
this scatter for the reasons given above). This scatter was to be expected due to the non-
steady situation in which the actual turbine was being tested (i.e. non-steady velocity and
reversing flow field).
234
Chapter 9
0.6
0.4
0.2
0.5 1.0 1.5 2.0 2.5
φ
Figure 9.4 Specific examples of instantaneous efficiency, η, as a function of flow factor,

φ, and the blade stagger angle of +/-45° during a number of wave peaks/troughs (length
of data record is 660s, power threshold is 10kW). Closed symbols are periods of
exhalation and open symbols are during inhalation.
2.5
2.0
1.5
*
P
1.0
0.5
0.0
0.5 1.0 1.5 2.0 2.5
φ
Closed symbols are periods of exhalation and open symbols are during inhalation.
235
Chapter 9
1.0
0.8
CT
0.5
0.3
0.0
0.5 1.0 1.5 2.0 2.5
φ
Closed symbols are periods of exhalation and open symbols are during inhalation.
Analysis of P* and CT coefficients (Figures 9.5 to 9.6) revealed that the field data
recorded during exhalations were shifted to higher values of φ compared to during
inhaling. This behaviour seems to be similar to the hysteresis effect that has been
reported by other researchers (e.g. Alcorn (2000), Setoguchi et al., (2003a), Thakker and
Abdulhadi (2008)).
Investigation of the causes of the hysteresis effects was beyond the scale of the
present study. However, there is certainly an important issue for future investigation in
regards to optimising the blade pitch control algorithm of the Denniss-Auld turbine so as
to obtain maximum efficiency from the Oceanlinx OWC plant. It should be noted that it
was apparent from a detailed examination of the experimental data that the response time
(or “damping”) of the STX2100 differential pressure transducer across the turbine was
somewhat longer than that of the other LP1000 pressure transducers. Thus, all data
presented in the current section has involved a time correction of 0.3s to the STZ2100
pressure readings.
The full set of processed experimental data (with an output power threshold of 10kW)
is presented in Figures 9.7 to 9.9. As in the case of γ = +/-45° it is seen that the
236
Chapter 9
instantaneous experimental data for other stagger angles had also noticeable scatter
(Figure 9.7).
0.8
70
60
0.6 55
50
η 45
0.4 -70
-60
-55
0.2
-50
-45
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Negative blade angle = inhaling φ
Figure 9.7 Instantaneous efficiency, η, as a function of flow factor, φ, and the blade
stagger angle, γ, (length of data record is 660s, power threshold is 10kW).
Analysis of η at a power threshold of 10kW showed that the most efficient operation
was achieved when the actual turbine blades were staggered at 60°.
237
Chapter 9
3 70
60
55
50
2
45
*
P -70
-60
1
-55
-50
-45
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
blade stagger angle, γ, (length of data record is 660s, power threshold is 10kW).
1.0
70
60
0.8
55
50
0.6 45
CT
-70
0.4 -60
-55
0.2 -50
-45
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
blade stagger angle, γ, (length of data record is 660s, power threshold is 10kW).
The analysis of the experimental data (with a power threshold of 10kW) provided
some important insights:
238
Chapter 9
• The most efficient operation of the full scale Denniss-Auld turbine was achieved
during exhalation wave cycles when γ = 60° and 1.7 ≤ φ ≤ 2.8.
• The widest operational range (1.7 ≤ φ ≤ 3.2.) during both flow cycles was also
demonstrated by the turbine with blades staggered at 60°.
• The peak instantaneous efficiency shown by the turbine with γ =60° was higher
by 9%, 11% and 8% than the peaks of η demonstrated by turbine with γ =45°, 50°
and 55°, respectively.
• The overall instantaneous efficiency of turbine with γ =60° during inhalation was
higher by approximately 10% than it was recorded when turbine operated with the
blades staggered at 70°.
• It would appear that the maximum power generated during the exhalation cycles
with the blades staggered at 70° did not exceed the 10kW threshold.
9.2 BEM analysis of the full scale Denniss -Auld turbine

A non-dimensional blade element analysis of a full scale Denniss-Auld turbine model
was one of the key tasks of the present study. As in the case of the lab-scale turbine the
major turbine performance parameters of this turbine model were simulated using a 10
streamtube BEM model in MatLab. The actual turbine had a hub-to-tip ratio of 0.77 as
depicted in Figure 9.10. It also had 21 cambered blades with variable stagger angle, γ, in
order to maintain pre-stall flow conditions (Figure 9.10). The overall rotor solidity, σ was
equal to 1.1.
The 10 streamtube model was implemented by dividing the span, b = (RT - RH), of
each blade into 10 sections. Analysis was performed through the range of flow
coefficient, φ, from 0.5 to 3.5 and using the stagger angles of 45°, 50°, 55°, 60° and 70°
similar to those in the experiments carried out on the actual turbine. Lift and drag
coefficients used in the analysis were deduced from earlier CFD modelling of the circular
arc profile isolated aerofoil used in the full scale Denniss-Auld turbine (provided in
Appendix A-3). The predicted performance characteristics of the full scale Denniss-Auld
turbine were compared against experimental data presented in the previous section.
239
Chapter 9
γ
RH
RT
Cambered
symmetrical about
the mid-chord blades
Figure 9.10 Geometrical parameters and technical features of the Denniss-Auld turbine
(made by present author).
9.2.1 Numerical data

Numerical results from the multiple streamtube BEM analysis of the full scale
Denniss-Auld turbine with a power threshold of 10kW, in terms of efficiency, η, pressure
coefficient, P*, and torque coefficient, CT, are presented in Figures 9.11 to 9.13.
Experimental full scale data described in the previous section are also given for
comparison. It should be mentioned that the numerical results presented below were
obtained without any special tuning of the MatLab simulations.
It is evident from Figure 9.11 that the non-dimensional 10 streamtube model
overestimated the turbine efficiency for all stagger angles tested in study. For example,
the closest prediction was made by the turbine model with γ = 60°, which overestimated
by 5% the peak efficiency recorded during exhalation and η data in the interval of 1.4 ≤ φ
≤ 2.0 for γ = 45° showed a trend similar to the predicted data, but the numerical results
were higher than the field data by approximately 10%. The most significant difference of
~34% between predicted and measured efficiency was shown by the model with the
largest stagger angle of +/-70º during inhalations.
240
Chapter 9
This overestimation of the efficiency by the BEM analysis was likely due to a number
of factors. Firstly, the multiple 10 streamtube non-dimensional BEM analysis was carried
without inclusion of a cascade interference factor. Secondly, the BEM analysis modelled
only the aerodynamic efficiency of the rotor in isolation. Losses in the drive train and
generator of the actual device were not included in the numerical results.
Analysis of P* data presented in Figure 9.12 showed that the field measurements
recorded during inhaling cycles, were predicted relatively accurately for all stagger
angles except 70º, where the simulation underestimated the magnitude of pressure
approximately by 50%. BEM results for non-dimensional torque, CT, demonstrated
reasonably good agreement to the experimental data obtained during inhaling cycles
(Figure 9.13). However, as in the previous case, CT exhalation data were noticeably
overestimated.
0.8 γ = 45º
70
60
55 70º
0.6
50
60º
η 45
55º
0.4 -70 50º
-60
-55
0.2 -50
-45
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Figure 9.11 Instantaneous efficiency (symbols), η, as a function of φ and γ compared to
the results based on the BEM analysis (solid lines).
241
Chapter 9
3 γ = 45º 50º 55º

70 60º
60
55
50
2
45
* -70
P
-60 70º
1 -55
-50
-45
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Figure 9.12 Instantaneous pressure coefficient (symbols), P*, as a function of φ and γ
compared to the results based on the BEM analysis (solid lines).
1.2
70
60 γ = 45º
50º
55
55º
0.8 50
60º
45
CT -70
-60
0.4 -55 70º
-50
-45
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Figure 9.13 Instantaneous torque coefficient (symbols), CT, as a function of φ and γ

compared to the results based on the BEM analysis (solid lines).
It is should be noted that the BEM model sensitivity to the number of streamtubes
used was tested and as expected it was found that similar numerical results (with a
242
Chapter 9
variation of ~ 1 %) were produced since this turbine case had a relatively large aspect
ratio, AR = 0.77.
9.3 CFD analysis of the full scale Denniss -Auld turbine

Results from the 3D CFD simulations of a full scale Denniss-Auld turbine and their
comparison with experimental data are presented in this section. Only a single blade
streamtube, i.e. 1/21st section of turbine, was modelled using the same boundary
conditions, GUIs and Frozen Rotor frame change as in case of the small scale CFD model
described in chapter 7. The results of the 3D CFD simulations presented in this section
were obtained on both the PC and the cluster computer. All CFD models were simulated
with tc =4.6mm, which was comparable with tip clearance of the actual turbine. Rotation
of the blades was modelled with an angular velocity of 36.6rad/s or 350RPM.

In order to determine the optimal number of structured elements for the
computational domain, grid independence tests were carried out before conducting the
main simulations. The structured elements were generated using TurboGrid and their
number was tested from 0.7×105 to 1.2×106. The tests showed that the optimum number
of elements, which did not substantially affect the numerical results, was ~1.0×106. Based
on this information, the final number of structured elements for a computational domain
was set to be between 1×106 and 1.1×106 depending on the value of the stagger angle.
9.3.2 Comparison with the field data

The experimental data discussed in section 9.1 were also used for comparison with
the 3D CFD results presented in Figures 9.14 to 9.17. Numerical results (broken lines)
deduced from the non-dimensional multiple streamtube BEM analysis described in the
previous section are also plotted for comparison. For better observation of numerical and
experimental efficiency data, the results are plotted on two graphs (Figures 9.14 and
9.15). It can be seen that the 3D CFD simulations noticeably improved the agreement
with the measured turbine efficiency compared to the results based on the BEM analysis.
The accuracy of prediction was improved by approximately 8%, 8%, 9%, 11% and 18%
for the stagger angles of 45°, 50°, 55°, 60° and 70°, respectively. It should be noted that
as in the case of BEM analysis, the CFD data present pure aerodynamic efficiency of the
turbine model. The losses in the drive train and generator were not included in the CFD
analysis.
243
Chapter 9
Analysis of pressure coefficients, P*, deduced from the 3D CFD simulations (Figure
9.16) revealed that the numerical pressure data (closed symbols/solid lines) were shifted
to lower values of φ compared to the data based on BEM analysis (broken lines). As a
result of that the agreement with the field data at lower values of φ was improved for all
stagger angles.
0.8 γ = 45º
70
55
45 55º
0.6
-70
η -55
0.4 70º
-45
0.2
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Figure 9.14 Instantaneous efficiency (symbols/broken lines), η, as a function of flow
the results of 3D CFD simulations (open symbols/solid lines) obtained for γ = 45º, 55º
and 70º. Curves from the BEM analysis are given for comparison (broken lines).
244
Chapter 9
0.8
60
50 γ = 50º
0.6 -60
-50
η 60º
50
0.4
0.2
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Figure 9.15. Instantaneous efficiency (symbols/broken lines), η, as a function of flow
the results of 3D CFD simulations (open symbols/solid lines) obtained for γ = 50º and
60º. Curves from the BEM analysis are given for comparison (broken lines).
4 70
60 γ = 45º
55
3
50 60º
* 45
50º 55º
P
2 -70
-60
-55 70º
1 -50
-45
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Figure 9.16 Instantaneous pressure coefficient (symbols), P*, as a function of flow factor,
φ, compared to the results of 3D CFD simulations (closed symbols/solid lines). Curves
from the BEM analysis are given for comparison (broken lines).
245
Chapter 9
1.2 70
γ = 45º
60
55
50º
0.8 50 55º
60º
45
CT -70
-60
0.4
-55 70º
-50
-45
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Figure 9.17 Instantaneous torque coefficient (symbols), CT, as a function of flow factor,
φ, compared to 3D CFD simulations (closed symbols/solid lines). Curves from the BEM
analysis are given for comparison (broken lines).
Analysis of the torque characteristics, CT, based on 3D CFD simulations

demonstrated a relatively close match to the BEM results for all stagger angles except 45°
where the CT data were shifted to significantly lower values of φ (see, Figure 9.17). The
fact that both sets of numerical data were matched closely to each other indicates that the
BEM simulations were not sensitive to the interference between blades due to the
“cascade” effect when γ ≥ 50°. It is interesting that differences between CFD and BEM
results for γ = 45° were larger than in other cases and this may be because CFD
fundamentally is capable of more realistic modelling of high solidity turbines than BEM.
9.4 Comparison of the full scale Denniss-Auld turbine with other

OWC turbines
A comparison of the performance of the full scale Denniss-Auld turbine (Figure 9.11)
with that of other self-rectifying turbines such as monoplane Wells turbines (Figure 2.22)
and Wells turbines with variable pitch blades (Figure 2.40) revealed some important
overall differences, as summarised in Table 9.1. It is seen from the evidence provided in
these figures that the Wells turbine with variable pitch blades potentially offers the most
246
Chapter 9
flexible range of flow factor φ (i.e. that the ratio of the highest to the lowest flow factors
in the turbine’s normal operating range is the largest of the three turbines). This turbine
also has the highest efficiency compared to the other turbines. In contrast to both Wells
turbines, the operational range of the Dennis-Auld turbine is shifted to much higher
values of φ, and its highest efficiency of 0.6 was achieved at φ = 2.0 when the stagger
angle was 60°. This is comparable with efficiency of the monoplane Wells turbine at φ =
0.11 reported by Curran and Gato (1997a). It is also interesting to note that the operating
efficiency of the Denniss-Auld turbine with blades staggered at 70° remained greater than
20% when the flow coefficient φ > 2.5. Comparisons of the non-dimensional pressure
drop characteristics of these three turbines (Figures 2.22, 7.7 and 9.12) also provides
some interesting results. Firstly, the P* curve of the monoplane Wells turbine
demonstrated the onset of stall at φ = 0.35. In contrast, both turbines with variable
geometries were able to operate without obvious stall. Moreover, both these turbines
showed a similar range of P* available by changing the blade stagger angle. However,
comparison of rotational speeds undoubtedly shows the Denniss-Auld turbine to be
preferable, as it has the lowest operable speed (~500RPM) compared to both Wells
turbines with typical speed of 2000 – 3000RPM.
Table 9.1 Summary of the performance characteristics of self-rectifying turbines

discussed in the present study.
Turbine Operational range (φ) ηmax
Monoplane Wells turbine (Figure 2.22) 0.05 – 0.2 0.6
Wells turbine with variable pitch blades
0.05 – 0.85 0.9
(Figure 2.40)
Denniss-Auld turbine (Figure 9.11) 1.0 – 3.0 0.6

Analysis of the experimental data obtained during sea trials of the entire OWC system
revealed that all instantaneous experimental data had noticeable scatter and hysteresis
effects. One of the critical issues in the processing of raw experimental data was
uncertainty in the determination of a true axial flow velocity used for evaluation of
turbine performance characteristics. It was found that with the power threshold of 10kW,
the most efficient operation of the full scale Denniss-Auld turbine was achieved during
247
Chapter 9
exhaling cycles when γ = 60° and 1.7 ≤ φ ≤ 2.8. This is comparable with efficiency of the
monoplane Wells turbine at φ = 0.11 reported by Curran and Gato (1997a).
The non-dimensional BEM analysis significantly overestimated the measured
efficiency of the full scale Denniss-Auld turbine. This could be due to a number of
factors including: a) the BEM analysis was based on the isolated lift/drag data while the
actual turbine had the high rotor solidity of 1.1; b) the losses in the drive train and
generator of the actual system could not be included in the numerical results.
Analysis of numerical data deduced from 3D CFD simulations of the full scale
Denniss-Auld turbine revealed that prediction of turbine efficiency was improved as
compared to that from the BEM analysis. It was also found that the CT data deduced from
CFD and BEM simulations closely match each other for all stagger angles except 45°
where the results may be sensitive to the blade interference and cascade effects.
Comparison of the full scale Denniss-Auld and the Wells turbine with variable pitch
blades showed a similar range of P* available by changing the blade stagger angle.
The Denniss-Auld turbine having the lowest operable speed (~500RPM) is more
preferable compared to both Wells turbines discussed in this chapter.
248
Chapter 10
CHAPTER 10 - CONCLUSIONS AND RECOMMENDATIONS

In this chapter the summary of the major outcomes obtained in the present study
followed by the recommendations for further research are presented.
10.1 Conclusions
• It has been demonstrate that accurate modelling of aerodynamic coefficients needed
for the design and analysis of turbomachinery can be achieved using CFD analysis
(viz. ANSYS CFX-11) providing that careful attention is paid to issues such as grid
refinement, boundary conditions, etc.
• CFD analysis carried out in this study confirmed the applicability of Weinig’s
analytical prediction of a cascade lift interference factor, k0, for the linear cascades of
NACA0012 and NACA0021 aerofoils at angles of attack less than 10°. Weinig’s
theory was based on an inviscid analysis conformal mapping analysis and predicted
that the cascade lift interference factor, k0, for an infinite cascade is independent of
the angle of attack, α, and is only a function of the stagger angle, γ, and ratio of s/c
(inverse solidity). However, in the present study it was found that the cascade
interference factors do indeed depend on the angle of incidence and therefore
Weinig’s inviscid model and equations such as (5.2) must be used with caution to
predict the lift of blades in a cascade at high angles of incidence (αm > 10º).
• Adjustment of CL and CD coefficients for cascade interference factors (k1 and δ1)
deduced from CFD simulations of the flow through aerofoils arranged in a linear
cascade significantly improved the potential of a non-dimensional model developed
in present study to predict the performance characteristics of a full scale Denniss-
Auld turbine.
• The non-dimensional multi-streamtube BEM analysis developed in the present work
was demonstrated to have the capability to predict the key performance
characteristics of OWC turbines with variable pitch angle blades with reasonable
accuracy.
• The study showed that the current design of the Denniss-Auld turbine can be further
improved so as to increase its operating efficiency by using CFD.
• It was found that the efficiency of the full scale Denniss-Auld turbine with variable
pitch blades was noticeably influenced by variation of tip clearance, tc; hub-to-tip
ratio, h, and number of blades, Ν. Adjustment of these parameters as follows: tc =
249
Chapter 10
2.3mm (or 1% of the length of c); h = 0.62; and N = 13 (or σ = 0.75) can increase
turbine efficiency by approximately 5.5% as compared to the baseline rotor
configuration. Additionally, it was shown that for the Denniss-Auld turbine, the
cambered blades of C type are preferable to the flat blades of type D. Besides this, it
was demonstrated that the baseline blade thickness of 0.06 provides an optimum
performance of a variable pitch turbine with blades of type C staggered at angles of
45º and 70º.
• Aerodynamic modelling of the full scale OWC wave energy converter comprising
major components such as a chamber, a nozzle-diffuser, a nacelle and a turbine with
variable pitch blades by means of CFD provided insights into the detailed behaviour
of the system. For example, it was found that the loss of pneumatic power in a
typical OWC system due to the presence of a 90° bend between the OWC chamber
and the horizontal turbine duct system was negligible (less than 0.01%).
• 3D CFD analysis revealed that the pneumatic efficiency of the OWC system based
on dimensions of an actual Oceanlinx wave energy plant to a large extent depends on
the instantaneous volume flow rate. It was determined that about half of the available
pneumatic power is lost in the OWC system as a result of frictional losses in the
nozzle-diffuser system when the volume flow rate, Q, is less than ~ 30m3/s.
• It was also found that a nacelle with conical nose cones was able to produce a more
uniform velocity field in the nozzle-diffuser upstream the turbine location for a
nacelle with rounded ends (baseline shape).
• 3D CFD simulations of the exhaling and inhaling airflows through the nozzle-
diffuser, nacelle and turbine with blades staggered at 45° showed that a uniform axial
velocity distribution was achievable upstream of the rotor. Moreover, it was found
that the 90º change of the airflow direction from the OWC chamber to the
converging nozzle section of the air duct did not distort the axial velocity field
approaching the rotor.
• 3D CFD analysis improved the prediction of the efficiency of a full scale Denniss-
Auld turbine by approximately 11% as compared to the non-dimensional BEM
analysis.
• Analysis of experimental data revealed that the full scale Denniss-Auld turbine has
operating efficiency comparable to the Wells turbines and its operational range is
shifted to much higher values of φ. Also due to lower rotational speed (~500RPM)
250
Chapter 10
the Denniss-Auld turbine is more preferable compared to the Wells turbines with
typical speed of 2000-3000RPM.
10.2 Recommendations for further research

During the present study some potential areas for further research were identified.
First, a fully transient CFD analysis of the Denniss-Auld turbine model incorporated in
the OWC converter should be carried out throughout the range flow conditions and
stagger angles typical for an actual turbine. This approach in contrast to the steady-state
simulations will provide a more realistic model of aerodynamic behaviour of the major
OWC components of the wave energy converter. Furthermore, it is desirable to apply
more realistic boundary conditions at the inlet of the turbine rather than of fixed axial
velocity as used in the present study. The model of air velocity imposed at the inlet (at
free water surface in OWC chamber) should be based on the realistic hydrodynamic
model of OWC motions. One of the major goals of further research should be to fully
couple the aerodynamic and hydrodynamic domains of the OWC system. In addition, it
would be desirable to incorporate modelling of the aerodynamic behaviour of the OWC
turbine with realistic mixtures of moist air and possibly seawater spray.
The present analysis of performance data obtained during the sea-trial tests of the
Oceanlinx OWC wave energy plant at Port Kembla were not able to definitively interpret
potential “hysteresis” effects or differences in performance of the turbine during exhaling
and inhaling and future studies should attempt to shed further light on this issue. This
may be an important task in order to obtain maximum efficiency from the OWC plant by
optimising the blade pitch control algorithm of the full scale Denniss-Auld turbine.
The CFD results presented in the present work were obtained simulating
incompressible flow. However, the actual operating conditions of full scale OWC
turbines are within a range that is very close to or even exceeds the critical Mach number.
In this respect, in future research it would be interesting to study the effect of Mach
number on the performance of the Denniss-Auld turbine as well as on overall
performance of the OWC system and its major components.
Last but not least potential area for further research mentioned in chapter 8 is CFD
analysis of aerodynamic efficiency of actual OWC system with inclusion of supporting
legs for turbine, generator and nacelle.
251
References
REFERENCES
ABBOTT, I. H. & VON DOENHOFF, A. E. (1959) Theory of wing sections (including a
summary of airfoil data), New York, Dover Publication, Inc.
ABBOTT, M. B. & BASCO, D. R. (1989) Computational Fluid Dynamics: An

Introduction for Engineers, New York, John Wiley & Sons, Inc.
ALCORN, R. (2000) Wave Station Modelling based on the Islay Prototype Plant. .
Belfast, UK, Queens University.
ALCORN, R. G. & FINNIGAN, T. D. (2004) Control Strategy Development for an

Inverter Controlled Wave Energy Plant. INTERNATIONAL CONFERENCE ON
RENEWABLE ENERGY AND POWER QUALITY (ICREPQ’04). Barcelona, Spain.
ANSYS (2010) http://www.ansys.com/assets/brochures/ansys-cfx-12.1.pdf. accessed:

09.01.2010
ARLITT, R. G. H., TEASE, K., STARZMANN, R. & LEES, J. (2007) Dynamic System
Modeling of an Oscillating Water Column Wave Power Plant based on Characteristic
Curves obtained by Computational Fluid Dynamics to enhance Engineered Reliability.
7th European Wave and Tidal Energy Conference. Porto, Portugal.
CASTRO, F., MARJANI, A., RODRIGUEZ, M. A. & PARRA, T. (2007) Viscous flow
analysis in a radial impulse turbine for OWC wave energy systems. 7th European Wave
and Tidal Energy Conference. Porto, Portugal.
CENGEL, Y. A. & CIMBALA, J. M. (2006) Fluid Mechanics Fundamentals and

Applications, Singapore, Mc Graw Hill.
CLEMENT, A., MCCULLEN, P., FALCAO, A., FIORENTINO, A., GARDNER, F.,
HAMMARLUND, K., LEMONIS, G., LEWIS, T., NIELSEN, K., PETRONCINI, S.,
PONTES, M.-T., SCHILD, P., SJOSTROM, B.-O., SORENSEN, H. C. & THROPE, T.
(2002) Wave energy in Europe: current status and prospectives Renewable and
Sustainable Energy Reviews, vol. 6, pp:405–431.
CRITZOS, C. C., HEYSON, H. H. & BOSWINKLE, R. W. (1955) Aerodynamic

characteristics of NACA 0012 airfoil section at angles of attack from 0° to 180°. 3361.
NACA, Technical Note 3361.
CRUZ, J. (2008) Ocean Wave Energy: Current Status and Future Prepectives, Berlin,
Springer-Verlag.
CURRAN, R. & GATO, L. M. (1997a) The energy conversion performance of several

types of Wells turbine designs. Power Energy, 211, Part A.
CURRAN, R., STEAWART, T. P. & WHITTAKER, T. J. T. (1997b) Design synthesis

of oscillating water column wave energy converters: performance matching. Mechanical
Engineering 21, 489-505.
252
References
CURRAN, R., WHITTAKER, T. J. T., RAGHUNATHAN, S. & BEATTIE, W. C.

(1998a) Performance Prediction of Contra-rotating Wells Turbines for Wave Energy
Converter Design. Journal of Energy Engineering, Vol.124, 35–53.
DAVEY, A. R. & TEASE, W. K. (1999) CFD Simulation of the Wells Turbine. 5th Int.
CFX Users Conference'99. Friedrichshafen.
DENNISS, T. (2007) A blade pitch control mechanism. WO 2007/009163 Al ,IN WIPO

(Ed. AUSTRALIAN PATENT OFFICE. Australia.
DHANASEKARAN, T. S. & GOVARDHAN, M. (2005) Computational analysis of
performance and flow investigation on Wells turbine for wave energy conversion.
Renewable Energy, 30, 2129–2147.
DIXON, S. L. (1978) Fluid Mechanics, Thermodynamics of Turbomachinery, Pergamon

Press, Oxford.
DORRELL, D. G. & HSIEH, M. F. (2008) Performance of Wells Turbines for Use in

Small-Scale Oscillating Water Columns. 18th International Offshore and Polar
Engineering Conference. Vancouver, BC, Canada.
DRESSER-RAND (2011) http://www.dresser-rand.com/products/hydroair/. accessed:

08/10/2011
EVANS, D. V. (1982) Wave-power absorption by systems of oscillating surface pressure

distributions. J. Fuid Mech, 114, 481-499.
EVANS, D. V., Ó GALLACHÓIR, B. P., PORTER, R. & THOMAS, G. P. (1995) On

the optimal design of an Oscillating Water Column device. 2nd European Wave Energy
Conf. Lisbon, Portugal.
EVES, A. (1986) The biplane Wells turbine. Belfast, Northern Ireland, The Queen's
University of Belfast.
FALCAO, A. F. D. O. (2000) The shoreline OWC wave power plant at the Azores. 4th
European Wave Energy Conference. Aalborg, Denmark.
FALCAO, A. F. D. O. & JUSTINO, P. A. P. (1999) OWC wave energy devices with air
flow control. Ocean Engineering, 26, 1275–1295.
FINNIGAN, T. (2004) Development of a 300kW ocean wave energy demonstration

plant. Pacific 2004 International Maritime Conference Sydney, NSW Australia
FINNIGAN, T. & ALCORN, R. G. (2003) Numerical simulation of a variable-pitch

turbine with speed control. 5th European Wave Energy Conference. Cork, Ireland.
FINNIGAN, T. & AULD, D. (2003) Model Testing of a Variable Pitch Aerodynamic

Turbine. 13th Int. Offshore and Polar Engineering Conf.
FUGLSANG, P., ANTONIOU, I., SØRENSEN, N. N. & MADSEN, H. A. (1998)

Validation of a Wind Tunnel Testing Facility for Blade Surface Pressure Measurements.
Roskilde, Risø National Laboratory.
253
References
GAREEV, A., COOPER, P. & KOSASIH, P. B. (2009) CFD Analysis of Air Turbines as
Power Take-Off Systems in Oscillating Water Column Wave Energy Conversion Plant.
8th European Wave and Tidal Energy Conference. Uppsala, Sweden.
GATO, L. M. C. & CURRAN, R. (1996) Performance of the Biplane Wells Turbine. J.

Offshore Mechanics and Arctic Engng., 118, 210-215.
GATO, L. M. C. & FALCAO, A. F. D. (1984) On the Theory of the Wells Turbine.

Journal of Engineering for Gas Turbines and Power, 106, 628-633.
GATO, L. M. C. & FALCAO, A. F. D. (1988) Aerodynamics of the Wells Turbine Int. J.

Mech. Sci, 30, 383-395.
GOVARDHAN, M. & CHAUHAN, V. S. (2007) Numerical studies on performance

improvement of self-rectifying air turbine for wave energy conversion. Engineering
Applications of Computational; Fluid Mechanics, Vol. 1, pp:57-60.
GRETTON, G. I. & BRUCE, T. (2007) Aspects of mathematical modelling of a

prototype scale vertical-axis turbine. 7th European Wave and Tidal Energy Conference.
Porto, Portugal.
HAWTHORNE, W. R. & HORLOCK, J. H. (1962) Actuator disc theory of the

incompressible flow in axial compressors. Proc Instn Mech Engrs, Vol. 176, pp: 789-
803.
HEATH, T., WHITTAKER, T. J. T. & BOAKE, C. B. (2000) The Design, Construction

and Operation of the LIMPET Wave Energy Converter (Islay, Scotland). 4th European
Wave Energy Conference. Denmark.
JACOBS, E. N. & SHERMAN, A. (1937) Aerofoil section characteristics affected by

variation of Reynolds number. NACA Report no. 586. NACA Report no. 586.
JAMSTEC (2009) Japan Marine Science and Technology Centre,

http://www.jamstec.go.jp/jamstec/myt.html. accessed: 01/09/2009
JONES, D. A. & CLARKE, D. B. (2003) An Evaluation of the FIDAP Computational

Fluid Dynamics Code for the Calculation of Hydrodynamic Forces on Underwater
Platforms DSTO-TR-1494 Fishermans Bend, Victoria 3207 Australia, Maritime
Platforms Division Platforms Sciences Laboratory
KANEKO, K., SETOGUCHI, T., HAMAKAWA, H. & INOUE, M. (1991) Biplane

Axial Turbine for Wave Power Generator. International Journal of Offshore and Polar
Engineering, Vol. 1, pp: 122-128.
KIM, T., KANEKO, K., SETOGUCHI, T. & INOUE, M. (1988) Aerodynamic

performance of an impulse turbine with self-pitch-controlled guide vanes for wave power
conversion. 1st KSME-JSME Thermal and Fluid Engineering Conference.
254
References
KIM, T., TAKAO, M., SETOGUCHI, T., KANEKOA, K. & INOUE, M. (2001)
Performance comparison of turbines for wave power conversion. Int. J. Therm. Sci., Vol.
40, pp: 681–689
KIM, T. H., SETOGUCHI, T., KANEKO, K. & RAGHUNATHAN, S. (2002) Numerical

investigation on the effect of blade sweep on the performance of Wells turbine.
Renewable Energy, Vol. 25, pp: 235–248.
LEE, C. H., NEWMAN, J. N. & NIELSEN, F. G. (1996) Wave interactions with an

oscillating water column. 6th International Offshore and Polar Engineering Conference.
Los Angeles.
LEISHMAN, J. G. (2000) Principles of Helicopter Aerodynamics, Cambridge University

Press.
LIMPET (2002) Islay LIMPET Wave Power Plant. Contract JOR3-CT98-0312. The
Queen’s University of Belfast.
LISSAMAN, P. B. (Ed.) (1998) Wind Turbine Aerofoils and Rotor Wakes, New York,
The American Society of Mechanical Engineers.
LOFTIN, L. K. & SMITH, M. A. (1949) Aerodynamic characteristics of 15 NACA

airfoil sections at seven Reynolds numbers. Technical Report 1945. NACA.
MAEDA, H., SANTHAKUMAR, S., SETOGUCHI, T., TAKAO, M., KINOUE, Y. &
KANEKO, K. (1999) Performance of an Impulse turbine with fixed guide vanes for wave
power conversion. Renewable Energy, Vol. 17, pp: 533-547.
MEI, C. C. (1976) Power Extraction from Water Waves. Journal of Ship Research, Vol.
20, pp: 63-66.
MENTER, F. R. (1992a) Performance of Popular Turbulence Models for Attached and

Separated Adverse Pressure Gradient Flow. AAIA Journal, Vol. 30, pp:2066-2072.
MENTER, F. R. (1994) Two-equation Eddy-viscosity Turbulence Model for Engineering

Applications. AAIA Journal, vol. 32, pp:1598-1605.
NEUMANN, F., BRITO-MELO, A., DIDIER, E. & SARMENTO, A. (2007) PicoOWC

Recovery Project: Recent Activities and Performance Data. 7th European Wave and
Tidal Energy Conference. Porto, Portugal.
NREC, C. (2011) http://www.conceptsnrec.com/News-And-

Events/Articles.aspx?page=5. accessed: 08/10/2011
OCEANENERGY (2011) http://www.oceanenergy.ie/index.html. accessed: 08/10/2011
PANDEY, M. M., PANDEY, K. P. & OJHA, T. P. (1988) Aerodynamic Characteristics

of Cambered Steel Plates in Relation to Their Use in Wind Energy Conversion Systems.
Wind Engineering, Vol. 12, pp: 90-104.
255
References
RAGHUNATHAN, S. (1988) Aerodynamic forces on airfoils at high angles of attack.

AIAA 1st National Fluid Dynamics Congress. Cincinnati, U.S.A.
RAGHUNATHAN, S. (1995) The Wells air turbine for wave energy conversion. Prog.
Aerosp. Sci., Vol. 31
pp: 335–386.
RAGHUNATHAN, S. (1996) Aerodynamics of Cascades at a Stagger Angle of 90

degrees. 34th Aerospace Sciences Meeting & Exhibit. Reno, NV, USA.
RAGHUNATHAN, S. & BEATTIE, W. C. (1996) Aerodynamic performance of contra-

rotating Wells turbine for wave energy conversion. IMechE, Vol. 210, pp: 431-447.
RAGHUNATHAN, S., SETOGUCHI, T. & KANEKO, K. (1990) Prediction of

Aerodynamic Performance of Wells Turbine From Aerofoil Data. Journal of
Turbomachinery, Vol. 112, pp: 792-795.
RAGHUNATHAN, S. & TAN, C. P. (1983) Aerodynamic performance of a Wells

turbine. Journal of Energy Engineering, Vol. 7, pp: 226-230.
RAGHUNATHAN, S., TAN, C. P. & OMBAKA, O. O. (1985) Performance of the Wells

self-rectifying air turbine. Aeronautical Journal, pp: 369-379.
SALTER, S. H. (1993) Variable Pitch Air Turbines. European Wave Energy Symposium.
Edinburgh.
SARMENTO, A. J. N. A. (1993) Model-test optimization of an OWC wave power plant.

Int J Offshore Polar Engng Vol.3, pp: 66–72.
SARMENTO, A. J. N. A. & FALCAO, A. F. D. O. (1985) Wave Generation by a

Oscillating Surface-pressure and its Application in Wave-energy Extraction. Journal of
Fluid Mechanics, Vol. 150, pp: 467-485.
SARMENTO, A. J. N. A., GATO, L. M. C. & FALCAO, A. F. O. (1990) Turbine-

controlled wave energy absorption by oscillating water column devices. Ocean Energy,
Vol.17, pp: 481-497.
SELIG, M. S., GUGLIELMO, J. J., BROEREN, A. P. & GIGUERE, P. (1995) Summary

of Low-Speed Airfoil Data, Virginia Beach, Virginia, SoarTech Publications.
SETOGUCHI, T., KANEKO, K., MAEDA, H., KIM, T. W. & INOUE, M. (1991)
Performance of impulse turbine with self-pitch-controlled guide vanes for wave power
conversion. 1st Int’l. Offshore and Polar Eng. Conf.
SETOGUCHI, T., KANEKO, K., TANIYAMA, H., MAEDA, H. & INOUE, M. (1996)
Impulse turbine with self-pitch-controlled guide vanes for wave power conversion: Guide
vanes connected by links. Int’l. Journal of Offshore and Polar Eng., Vol. 6, pp: 76-80.
SETOGUCHI, T., KIM, T. H., KANEKO, K., TAKAO, M., LEE, Y. W. & INOUE, M.
(2002) Air Turbine with Staggered Blades for Wave Power Conversion. 12th
International Offshore and Polar Engineering Conference. Kitakyushu, Japan, The
International Society of Offshore and Polar Engineers.
256
References
SETOGUCHI, T., KINOUE, Y., KIM, T. H., KANEKO, K. & INOUE, M. (2003)
Hysteretic characteristics of Wells turbine for wave power conversion. Renewable
Energy, Vol. 28, pp: 2113–2127.
SETOGUCHI, T., RAGHUNATHAN, S., TAKAO, M. & KANEKO, K. (1997) Air-

Turbine with Self-Pitch-Controlled Blades for Wave Energy Conversion (Estimation of
Performances in Periodically Oscillating Flow). International Journal of Rotating
Machinery Vol.3, pp: 233-238.
SETOGUCHI, T., SANTHAKUMAR, S., TAKAO, M., KIM, T. H. & KANEKO, K.

(2003b) A modified Wells turbine for wave energy conversion. Renewable Energy, Vol.
28, pp: 79–91.
SETOGUCHI, T., TOKAO, M., ITAKURA, K., MOHAMMAD, M., KANEKO, K. &
THAKKER, A. (2003a) Effect of Rotor Geometry on the Performance of the Wells
Turbine. 13th International Offshore and Polar Engineering Conference Honolulu,
Hawaii, USA.
SHAW, C. T. (1992) Using Computational Fluid Dynamics, New York, Prentice Hall.
SHELDAHL, R. E. & KLIMAS, P. (1981) Aerodynamic characteristics of Seven

Symmetrical Airfoil Sections Through 180-Degree Angle of Attack for Use in
Aerodynamic analysis of Vertical Axis Wind Turbines. SAND80-2114.
TAKAO, M., ITAKURA, K., SETAGUCHI, T., KIM, T. H., KANEKO, K. &
THAKKER, A. (2002) Performance of a Radial Turbine for Wave Power Conversion.
12th International Offshore and Polar Engineering Conference. Kitakyushu Japan.
TAKAO, M., SETOGUCHI, T., KANEKO, K. & INOUE, M. (1997) Air Turbine with
Self-Pitch Controlled Blades for Wave Energy Conversion. International Journal of
Offshore and Polar Engineering, Vol. 7.
TAKAO, M., THAKKER, A., ABDULHADI, R. & SETOGUCHI, T. (2006) Effect of

blade profile on the performance of a large-scale Wells turbine for wave-energy
conversion. International Journal of Sustainable Energy, Vol. 25, pp: 53-61.
TAYLOR, J. R. M. & CALDWELL, N. J. (1998) Design and Construction of the

Variable-Pitch Air Turbine for the Azores Wave Energy Plant. 3rd European Wave
Energy Conference. Patras, Greece.
TEASE, W. K. (2003) Dynamic Response of a Variable Pitch Wells Turbine. 5th

European Wave Energy Conf. Cork, Ireland.
TEASE, W. K., LEES, J. & HALL, A. (2007) Advances in Oscillating Water Column Air
Turbine Development. 7th European Wave and Tidal Energy Conference. Porto,
Portugal.
THAKKER, A. & ABDULHADI, R. (2007) Effect of Blade Profile on the Performance

of Wells Turbine under Unidirectional Sinusoidal and Real Sea Flow Conditions.
International Journal of Rotating Machinery, Vol. 2007.
257
References
THAKKER, A. & ABDULHADI, R. (2008) The performance of Wells turbine under bi-
directional airflow. Renewable Energy 33, 2467–2474.
THAKKER, A. & DHANASEKARAN, T. S. (2003) Computed effects of tip clearance

on performance of impulse turbine for wave energy conversion. Renewable Energy, Vol.
29, pp: 529–547.
THAKKER, A. & ELHEMRY, M. A. (2007) 3-D CFD ANALYSIS ON EFFECT OF

HUB-TO-TIP RATIO ON PERFORMANCE OF IMPULSE TURBINE FOR WAVE
ENERGY CONVERSION. THERMAL SCIENCE. , Vol.11, pp:157-170.
THAKKER, A., FRAWLEY, P. & KHALEEQ, H. B. (2002) An Investigation of the

Effects of Reynolds Number on the Performance of 0.6 m Impulse Turbines for Different
Hub to Tip Ratios. 12th International Offshore and Polar Engineering Conference.
Kitakyushu, Japan, Th e International Society of Offshore and Polar Engineers.
THAKKER, A., FRAWLEY, P., KHALEEQ, H. B., ABUGIHALIA, Y. &

SETOGUCHI, T. (2001b) Experimental and CFD analysis of 0.6 m Impulse turbine with
fixed guide vanes. 11th International Offshore and Polar Engineering Conference.
Stavanger, Norway.
THAKKER, A., FRAWLEY, P. & SHEIK BAJEET, E. (2001a) Numerical analysis of

Wells turbine performance using a 3D Navier–Strokes explicit solver. 11th ISOPE.
Stavanger, Norway.
THAKKER, A., HOURIGAN, F., DHANASEKARAN, T. S., HEMRY, M. E.,

USMANI, Z. & RYAN, J. (2005a) Design and performance analysis of impulse turbine
for a wave energy power plant. INTERNATIONAL JOURNAL OF ENERGY RESEARCH,
Vol. 29, pp: 13-36.
THIEBAUT, F., O‘SULLIVAN, D., KRACHT, P., CEBALLOS, S., LÓPEZ, J.,
BOAKE, C., BARD, J., BRINQUETE, N., VARANDAS, J., GATO, L. M. C.,
ALCORN, R. & LEWIS, A. W. (2011) Testing of a floating OWC device with movable
guide vane impulse turbine power take-off. 9th European Wave and Tidal Energy
Conference (EWTEC 2011). University of Southampton, UK, 5–9 September 2011
TORRESI, M., CAMPOREALE, S. M. & PASCAZIO, G. (2007) Experimental and

numerical investigation on the performance of a Wells turbine prototype. 7th European
Wave and Tidal Energy Conference
Porto, Portugal.
TORRESI, M., CAMPOREALE, S. M., STRIPPOLI, P. D. & PASCAZIO, G. (2008)

Accurate numerical simulation of a high solidity Wells turbine. Renewable Energy,
Vol.33, pp:735-747.
TU, J., YEOH, G. H. & LIU, C. (2008) Computational Fluid Dynamics: A Practical
Approach, Amsterdam, ELSEVIER.
TURTON, R. K. (1984) Principles of turbomachinery E. & F.N. Spon.
258
References
VERSTEEG, H. K. & MALALASEKERA, W. (2007) An Introduction to

COMPUTATIONAL FLUID DYNAMICS (The Finite Volume Method), Glasgow, Pearson
Education Limited.
WALLIS, R. A. (1961) Axial Flow Fans (Design and Practice), New York.
WALLIS, R. A. (1983) Axial Flow Fans and Ducts, London, John Willey & Sons.
WATTERSON, J. K. & RAGHUNATHAN, S. (1996) Investigation of Wells turbine

performance using 3-D CFD. Energy Conversion Engineering Conference. Washington,
DC, USA.
WATTERSON, J. K. & RAGHUNATHAN, S. (1998) Computed effects of solidity on

Wells turbine performance. JSME Int J Series B, Vol. 41, pp: 199–205.
WAVENET (2003) Energy, Environment and Sustainable Development Program.
WEBER, J. W. (1999) System Analysis for an Oscillating Water Column wave energy
device. Marie Curie Grand Holder Meeting, Almeria, Spain.
WEBER, J. W. & THOMAS, G. P. (2000) Optimisation of the hydrodynamic-

aerodynamic coupling for an Oscillating Water Column wave energy device. 4th
European wave Energy Conf. Aalborg, Denmark
WEBER, J. W. & THOMAS, G. P. (2001) An investigation into the importance of the air
chamber design of an Oscillating Water Column wave energy device. ISOPE 2001.
Stavanger, Norway.
WEC (2007) 2007 Survey of Energy Resources: Executive Summary. World Energy
Council.
WEINIG, F. S. (1964):Theory of two-dimensional flow through cascades. Section B (in

Aerodynamics of turbines and compressors), Princeton University Press.
WHITAKER, S. & PIGFORD, R. L. (1960) Numerical Differentiation of Experimental

Data. Industrial and Engineering Chemistry, Vol. 52, pp:185 -187.
WIKIPEDIA (2011) http://en.wikipedia.org/wiki/Hysteresis. accessed: 07/10/2011
WILCOX, D. C. (1988) Reassessment of the Scale-determining Equation for Advanced

Turbulence Model. AAIA Journal, vol. 26, pp:1299 -1310.
WONG, D. S. L. (1994) Prediction of the Performance of the Wells Turbine. Department

of Aeronautical Engineering. Belfast, The Queen's University of Belfast.
YILBAS, B. S., BUDAIR, M. O. & AHMED, M. N. (1998) Numerical simulation of the

flow field around a cascade of NACA0012 airfoils-effects of solidity and stagger.
Comput. Methods Appl. Mech. Engrg., Vol. 158, pp: 143-154.
259
Appendixes
Appendix A-1
Before conducting the major CFD simulations in accordance with the research
objectives it is essential to assess the CFD methodology adopted in the present study. Of
particular importance is to test the capability of the turbulence models available in
ANSYS CFX-11 to provide accurate prediction of the real fluid flows. To accomplish
this, a CFD model of air flow through an orifice integrated in a pipe was compared with
the empirical solution. The situation chosen involves an incompressible steady flow of air
in a horizontal pipe of diameter, D, that is constricted to a flow area of diameter d as
shown in Figure A.1. Using the mass balance and the Bernoulli equation between a
location before the constriction (point 1) and the location of the orifice (point 2), the
pressure drop across the orifice can be expressed as:
(P1 − P2 ) = ∆P = (
V12 ρ 1 − ϕ 4
,
) (A.1)
2Cd2ϕ 4
where V1 is the average velocity through the cross sectional area in the point 1, ρ is the
fluid density, φ=d/D is the diameter ratio, and Cd is the discharge coefficient, which takes
into account the losses due to the obstruction
The value of Cd depends on both φ and Re and for the flow condition with Re > 3×104
and for orifices can be taken as 0.61 (Cengel and Cimbala, 2006). Note: the model
presented in Figure 3.18 was based on Re = 4.7×104..
Obstruction
1 D 2 d
Figure A. 1 Flow through a constriction in a pipe.
Assuming D=0.04m, d=0.02m, V1=10m/s, air density ρ=1,185 kg/m3, and Cd=0.61,
equation (A.1) gave ∆P=2388.47Pa. A CFD model was then built using these pipe/orifice
dimensions. The length of the pipe in CFD model was taken as 0.5m; the orifice has
been placed in the middle of the pipe. The normal speed of 10m/s with turbulence
intensity of 5% was imposed at the inlet, a static pressure of 0Pa was specified at the
outlet, a no-slip wall condition was taken for the inside pipe surface and orifice. In order
to account for the viscous flow in the near-wall region, the 12 inflated layers have been
260
Appendixes
generated on the inside pipe surface and orifice. The height of the first inflation layer was
taken as 0.1mm, and the expansion factor as 1.1. The computational domain was meshed
using unstructured elements. For meshing of a high gradient region at the location of the
orifice (see, Figure A.2), “point control” with adjustable spacing parameters was used at
the centre of the orifice (x=0, y=0, and z=0) as shown in Figure A.3. The k-ε turbulence
model with the scalable wall function was used for the flow simulations.
Orifice
Inlet Outlet
Figure A. 2 Velocity countor plot in vicinity of the orifice.
X
Z
Figure A. 3 Adjustment of the size of mesh in the vicinity of the orifice by applying
“point control”.
The final CFD solution for comparison against the analytical model was based on the
grid independence testing. The test results in terms of ∆P across the orifice are presented
in Figure A.4. The grid independence testing revealed that an increase in the number of
grid elements, n, more than 1.42×106 did not change the value of ∆P appreciably. The
pressure drop deduced from CFD simulation with n = 1.42×106 was equal to 2381.8Pa.
This numerical value of ∆P was lower by 0.28% the empirical result (2388.47Pa).
261
Appendixes
2.4
2.2
∆P (kPa)
1.8
1.6
0.2 0.6 1 1.4 1.8 2.2
n ×106
Figure A. 4 Results of grid independence testing.
Based on assessment given above it can be concluded that the CFD methodology
adopted for the study is capable of providing accurate prediction of empirical solution
using the k-ε turbulence model and comparable value of Re.
262
Appendixes
Appendix A-2
Angle of Lift Coefficient, CL Drag Coefficient, CD

attack
(deg) Re=1×105 Re=3×105 Re=6×105 Re=1×105 Re=3×105 Re=6×105
-90 -0.002 0.000 -0.001 0.768 1.316 1.341

-80 -0.128 -0.174 -0.190 1.231 1.262 1.326
-75 -0.176 -0.242 -0.262 1.177 1.231 1.299
-70 -0.246 -0.305 -0.319 1.142 1.169 1.213
-65 -0.331 -0.384 -0.404 1.088 1.105 1.138
-60 -0.349 -0.454 -0.485 1.007 1.042 1.065
-55 -0.287 -0.409 -0.449 0.819 0.890 0.928
-50 -0.293 -0.378 -0.403 0.706 0.738 0.752
-45 -0.263 -0.369 -0.401 0.600 0.640 0.655
-40 -0.207 -0.299 -0.324 0.491 0.519 0.527
-35 -0.144 -0.244 -0.272 0.398 0.426 0.434
-30 -0.107 -0.168 -0.195 0.341 0.342 0.346
-25 -0.028 -0.100 -0.126 0.253 0.258 0.265
-20 0.110 0.031 0.013 0.173 0.181 0.182
-15 0.096 0.164 0.139 0.146 0.122 0.123
-10 0.136 0.203 0.111 0.085 0.080 0.087
-5 -0.289 -0.128 0.086 0.077 0.060 0.041
-3 0.238 0.193 0.207 0.031 0.028 0.027
0 0.485 0.454 0.465 0.024 0.021 0.019
3 0.751 0.739 0.758 0.024 0.021 0.019
5 0.933 0.925 0.940 0.026 0.022 0.021
8 1.145 1.151 1.167 0.034 0.029 0.027
10 1.269 1.285 1.301 0.043 0.037 0.033
13 1.360 1.415 1.421 0.064 0.055 0.045
15 1.362 1.415 1.442 0.085 0.072 0.061
20 1.187 1.211 1.280 0.171 0.156 0.142
25 0.955 0.957 0.959 0.293 0.296 0.293
30 0.899 0.862 0.859 0.459 0.458 0.464
40 1.238 1.234 1.271 0.910 0.898 0.939
50 1.231 1.210 1.190 1.377 1.305 1.288
60 0.763 0.884 0.938 1.558 1.655 1.676
70 0.665 0.631 0.655 1.763 1.712 1.768
80 0.324 0.267 0.297 1.847 1.518 1.810
90 0.000 0.002 0.001 1.103 1.046 1.014
263
Appendixes
Appendix A-3
Angle of Lift Coefficient, CL Drag Coefficient, CD

attack
(deg)
Re=2.5×105 Re=5×105 Re=7.5×105 Re=1×106 Re=2.5×105 Re=5×105 Re=7.5×105 Re=1×106
-90 0.001 -0.001 -0.001 -0.001 0.986 0.969 0.975 0.961

-80 -0.171 -0.170 -0.263 -0.254 1.141 1.128 1.663 1.627
-70 -0.450 -0.477 -0.481 -0.475 1.381 1.458 1.463 1.462
-60 -0.711 -0.727 -0.738 -0.749 1.365 1.366 1.380 1.390
-55 -0.601 -0.592 -0.596 -0.650 1.110 1.095 1.089 1.123
-50 -0.678 -0.695 -0.713 -0.755 0.956 0.964 0.968 1.000
-45 -0.767 -0.775 -0.810 -0.812 0.899 0.894 0.917 0.914
-40 -0.806 -0.814 -0.813 -0.822 0.788 0.781 0.776 0.778
-35 -0.354 -0.385 -0.363 -0.383 0.458 0.479 0.472 0.465
-30 -0.264 -0.359 -0.436 -0.397 0.369 0.410 0.433 0.414
-25 -0.186 -0.220 -0.242 -0.291 0.258 0.276 0.281 0.300
-20 -0.254 -0.151 -0.157 -0.126 0.198 0.182 0.183 0.182
-15 0.033 -0.056 -0.112 -0.132 0.112 0.128 0.126 0.129
-10 -0.133 -0.122 -0.098 -0.086 0.090 0.089 0.088 0.087
-5 0.038 0.029 0.030 0.033 0.026 0.024 0.023 0.023
-3 0.220 0.218 0.220 0.221 0.026 0.024 0.024 0.024
0 0.488 0.505 0.512 0.502 0.021 0.019 0.018 0.016
3 0.717 0.737 0.748 0.751 0.021 0.020 0.019 0.019
5 0.870 0.895 0.903 0.909 0.025 0.023 0.023 0.022
8 1.066 1.092 1.103 1.109 0.034 0.032 0.031 0.031
10 1.161 1.182 1.191 1.195 0.044 0.043 0.043 0.042
13 1.168 1.181 1.187 1.187 0.069 0.068 0.067 0.067
15 1.093 1.075 1.051 1.043 0.105 0.108 0.110 0.109
20 0.933 0.741 0.509 0.508 0.262 0.242 0.225 0.247
25 0.854 0.887 0.889 0.922 0.357 0.380 0.399 0.428
30 0.825 0.844 0.863 0.870 0.516 0.567 0.580 0.577
40 0.973 0.902 0.882 0.903 0.885 0.896 0.884 0.892
50 0.967 0.913 0.923 0.816 1.338 1.313 1.264 1.225
60 0.831 0.867 0.847 0.863 1.561 1.553 1.573 1.622
70 0.633 0.659 0.660 0.644 1.732 1.775 1.784 1.759
80 0.358 0.367 0.368 0.356 1.943 1.949 1.961 1.926
90 -0.026 -0.027 -0.027 -0.028 1.296 1.261 1.244 1.245
264
Appendixes
Appendix A-4
265
Appendixes
Appendix A-5
Boundary conditions and settings for CFD modelling of an isolated aerofoil
Computational domain (free-slip wall)
Inlet Outlet
(Normal speed) Aerofoil (no-slip wall)
(Pst = 0Pa)
Turbulence
Intencity-5%
266
Appendixes
Appendix A-6
Values of lift interference factor, k0, deduced from the CFD simulations in present study.
Stagger angle, γ′
s/c
0° 20° 30° 40° 60° 70° 80°
0.5 0.26 1.03
0.8 0.38 1.07 1.37 2.06
1 0.46 1.00 1.22 1.52 2.81 3.71
1.3 0.57 0.99 1.10 1.25 1.66 1.85 1.89
1.5 0.64 0.99 1.07 1.18 1.42 1.51 1.50
2 0.79 0.96 1.01 1.06 1.17 1.22 1.22
2.5 0.93 0.95 0.98 1.02 1.15 1.12 1.13
4 0.99 1.00 1.00 1.00 1.03 1.04 1.06
Values of lift interference factor, δ0, deduced from the CFD simulations in present study.
Stagger angle, γ′
s/c
0° 20° 30° 40° 60° 70° 80°
0.5 2.36 1.85
0.8 2.59 6.15 7.32 7.72
1 2.60 1.28 1.37 1.48 1.05 1.17
1.3 2.49 1.12 1.14 1.22 1.31 1.30 1.26
1.5 2.29 1.07 1.10 1.12 1.16 1.15 1.14
2 1.69 1.03 1.07 1.10 1.07 1.08 1.01
2.5 0.94 1.00 1.02 1.05 0.98 1.00 0.97
4 0.89 0.95 0.98 1.02 1.01 0.98 0.93
267
Appendixes
Appendix A-7
Figures A.5 to A.7 show the pressure contours at the tip and the leading edge of the
turbine blade with different values of tc staggered at 45°, 55° and 70°, respectively. It can
be seen that the pressure at the tip of the leading edge varied depending on a value of tc
for all stagger angles. Variations of pressure from inlet to outlet with blades staggered at
45°, 55º and 70º and different values of tc (TGs) are shown in Figures A.8 to A.10,
respectively. These data complement the data presented in Figures A.5 to A.7. It is seen
from Figure A. 8 that the turbine model without tip clearance (TG=0%) demonstrated the
highest pressure at the inlet. The models with tip clearance produced the lower pressures
at the inlet. Analysis showed that the pressure drop across the rotor, ∆P, with blades
staggered at 45º and having TG=1%, TG=2% and TG=3% was decreased by 3%, 6% and
10% compared to the model without tip clearance, respectively. It also can be seen from
Figures A.9 to A.10 that the models with γ = 55º and γ = 70º, which were tested with
TG=1% were less sensitive to variation of inlet pressure than configurations with larger
values of tc. For instance, the inlet pressure of the rotors with γ = 55º and γ = 70º was
reduced only by 1% and less 1% respectively compared to the turbine configuration
without tc. Larger values of tc provided more noticeable reduction of the inlet pressure
and consequently the pressure drop across the rotor. The model of the full scale Denniss-
Auld turbine with γ = 55º demonstrated decrease of ∆P by 5% and 8% at larger values of
TG=2% and TG=3%, respectively compared to the rotor configuration without tc. The
same trend was shown by the model with γ = 70º, where ∆P was decreased by 4% and by
8% with increase of tc from 2% to 3%.
268
Appendixes
tc= 0% tc= 1%
γ = 45°, φ = 1.18
tc= 2% tc= 3%
Figure A. 5 Pressure contours at the tip and the leading edge of the turbine blade with
different values of tc staggered at 45º. Flow coefficient, φ, = 1.18.
tc= 0% tc= 1%
γ = 55°, φ = 1.57
tc= 2% tc= 3%
269
Appendixes
tc= 0% tc= 1%
γ = 70°, φ = 2.36
tc= 2% tc= 3%
TG=0%
102000
TG=1%
TG=2%
101800 TG=3%
Pressure (Pa)
101600
101400 Patm
101200
101000
0 1 2 3
Streamwise location (m)
Figure A. 8 Pressure variations from inlet to outlet of a model of Denniss-Auld turbine

with γ = 45º and different values of tc at φ = 1.18 (complements Figure A.5).
270
Appendixes
TG=0%
102000
TG=1%
TG=2%
Pressure (Pa) 101800 TG=3%
101600
101400 Patm
101200
101000
0 1 2 3

TG=0%
101800
TG=1%
TG=2%
TG=3%
101600
Pressure (Pa)
101400
Patm
101200
101000
0 1 2 3

271
Appendixes
Appendix A-8
Data deduced from 3D CFD simulations of a turbine with γ = 55º are presented in
Figures A.11 to A. 13. As in the case of γ = 45º, the turbine configuration with the larger
stagger angle of 55º demonstrated a similar trend. At low flow coefficients, φ, the
baseline model with the highest solidity, σ = 1.11 was more efficient than the models
with lower values of σ (Figure A.11). At higher flow coefficients (φ > 1.57) decrease of
σ provided higher efficiency of the turbine models. It can be seen that the critical value of
σ was 0.47, the efficiency provided by the model with this solidity was approximately by
2% lower compared to the model with σ = 0.58 through the entire range of φ. Analysis of
all η data indicated that the optimum value of solidity can be 0.68. In the most efficient
operational range of 1.75 < φ < 2.75, the model with this solidity increased turbine
efficiency approximately by 1% and 3% compared to the models with σ = 0.9 and σ =
1.11 respectively. However, at lower values of 1.25 < φ < 1.75 the model with σ = 0.68
performed approximately by 3% less efficiently than the model with higher solidity of
0.9. However, at the lowest value of φ = 1.18 the model with σ = 0.68 provided 9% more
efficient operation than it was demonstrated by the turbine configuration with σ = 0.9. It
can be seen that the effect of σ on torque coefficient, CT, was less noticeable at lower
values of φ up to 1.57 (Figure A.12). At higher φ > 1.57, the torque characteristics of the
model with γ = 55° were visibly increased with decrease of σ. In the interval of 1.57 < φ
< 2.75, the highest CT data were shown by the model with the lowest value of σ. In the
same interval of φ, the coefficients of CT produced by the model with σ = 0.68 were
approximately by 6% and 12% higher compared to the torque characteristics
demonstrated by the models with σ = 0.9 and σ = 1.11, respectively. The coefficients of
P* deduced from 3D CFD simulations of the turbine with blades staggered at 55°
demonstrated the similar trend as the P* data based on the model with γ = 45° (Figure
A.13). It can be seen the coefficients of P* were significantly reduced by a decrease of
σ throughout the range of φ tested in study. For instance, the P* data based on the model
with the lowest σ = 0.47 were approximately by 49% lower compared to the pressure
data based on the baseline configuration (σ = 1.11).
272
Appendixes
0.8
σ = 1.11
γ = 55º
σ = 0.9
0.6 σ = 0.68
σ = 0.58
η σ = 0.47
0.4
0.2
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
φ
Figure A. 11 Effect of rotor solidity, σ, on efficiency, η, of the turbine model with blades
staggered at 55º.
1.2
σ = 1.11 γ = 55º
σ = 0.9
σ = 0.68
0.8
σ = 0.58
CT σ = 0.47
0.4
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
φ
Figure A. 12 Effect of rotor solidity, σ, on torque characteristic, CT, of the turbine model
273
Appendixes
3 γ = 55º
σ = 1.11
σ = 0.9
σ = 0.68
2 σ = 0.58
σ = 0.47
*
P
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
φ
Figure A. 13 Effect of rotor solidity, σ, on pressure coefficient, P*, of the turbine model
Results in terms of η, CT, and P* deduced from 3D CFD simulations of a turbine with
γ = 70º are presented in Figures A.14 to A.16. It is clearly seen from Figure A.14 that
when the turbine was modelled with the largest stagger angle of 70º, the higher rotor
solidity provided the higher efficiency. For instance, the baseline turbine configuration (σ
= 1.11) was approximately by 3% more efficient compared to the model with solidity of
0.9. Analysis of CT data showed that at lower flow coefficients, φ, the rotors with higher
solidities provided the higher values of CT (Figure A.15). However, in the interval of φ >
2.55, the turbine models with largest stagger angle of 70º demonstrated the opposite
behaviour. The rotors with lower solidities produced higher torque characteristics.
Analysis of P* data presented in Figure A.16 revealed the same trend as was shown by
the models with smaller stagger angles (45º and 55º). The rotors with higher solidities
demonstrated higher pressure coefficients, P*. For example, the P*data deduced from 3D
CFD simulations of a baseline turbine model (σ = 1.11) were approximately by 15%
higher than it was shown by the model with lower solidity of 0.9. In contrast to the
models with small stagger angles, the turbine configuration with γ = 70° demonstrated
more linear distribution of P* for all rotor solidities through the range of φ tested in study.
274
Appendixes
0.5
γ =70º
σ = 1.11
0.4 σ = 0.9
σ = 0.68
0.3 σ = 0.58
η
0.2
0.1
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
φ
Figure A. 14 Effect of rotor solidity, σ, on efficiency, η, of the turbine model with blades
staggered at 70º.
0.4
σ = 1.11 γ =70º
σ = 0.9
0.3
σ = 0.68
CT σ = 0.58
0.2
0.1
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
φ
Figure A. 15 Effect of rotor solidity, σ, on torque characteristic, CT, of the turbine model
275
Appendixes
1.2 γ =70º
σ = 1.11
σ = 0.9
σ = 0.68
0.8
σ = 0.58
*
P
0.4
0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
φ
Figure A. 16 Effect of rotor solidity, σ, on pressure coefficient, P*, of the turbine model
276

Analysis of Variable Pitch Air Turbines For Oscillating Water Col PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Analysis of Variable Pitch Air Turbines For Oscillating Water Col PDF

Uploaded by

Copyright:

Available Formats

University of Wollongong

Analysis of variable pitch air turbines for oscillating

Research Online is the open access institutional repository for the

A thesis submitted in partial fulfilment of the requirements

SCHOOL OF MECHANICAL, MATERIALS AND MECHATRONIC

“__” ______ 2011

range and technical reliability.

Momentum (BEM) method revealed a number of “cascade” issues having particular

performance of OWC axial flow turbines using only generic/non-dimensional input

ratio, h, and number of blades, Ν. Adjustment of these design parameters so that

tc = 2.3mm, h = 0.62, and N = 13 resulted in a predicted increase in turbine efficiency by

up 5.5% compared to the actual/baseline rotor configuration. CFD modelling to ascertain

the influence of the non-dimensional thickness of the blades in a Denniss-Auld turbine

components of a turbine/OWC system, such as the nozzle-diffuser, nacelle and rotor. It

appeared to have no detrimental effect on the uniformity of the axial velocity as a

function of circumferential position around the turbine.

by approximately 11% as compared to the BEM method.

to acknowledge the assistance, guidance and support of the following people:

• Doctor Brad Stappenbelt, who gave me valuable advice and encouragement

during my final year.

• Oceanlinx for provision of experimental data on the performance of their full

scale turbine and OWC system at Port Kembla.

• Mr Patrick McGowan who listened and encouraged me throughout my project.

TABLE OF CONTENTS ........................................................................................................................... VI

TABLE OF FIGURES ............................................................................................................................... XI

NOMENCLATURE ............................................................................................................................. XXVI

1.1 MOTIVATION AND OBJECTIVES ......................................................................................................3

CHAPTER 2 - DEVELOPMENT OF OWC TURBINES..................................................................7

2.1 OWC TECHNOLOGY AS A PRACTICAL MEANS OF WAVE ENERGY CONVERSION ..............................7

CHAPTER 3 - METHODOLOGIES FOR AERODYNAMIC ANALYSIS OF OWC AXIAL

3.1 BLADE ELEMENT MOMENTUM (BEM) ANALYSIS.........................................................................74

CHAPTER 4 - CFD ANALYSIS OF ISOLATED AEROFOILS ..................................................100

4.1 EXPERIMENTAL CL AND CD COEFFICIENTS FOR VARIOUS AEROFOIL PROFILES...........................100

CHAPTER 5 - TWO DIMENSIONAL LINEAR CASCADES AND DETERMINATION OF

5.1 ANALYTICAL PREDICTION OF CASCADE INTERFERENCE ............................................................130

CHAPTER 6 - DEVELOPMENT OF A NON-DIMENSIONAL FORM OF THE BEM

6.1 NON-DIMENSIONAL MULTIPLE STREAMTUBE BEM ANALYSIS ...................................................157

CHAPTER 7 - CFD ANALYSIS OF AXIAL FLOW TURBINES................................................170

7.2.3 Modelling of the full turbine coupled to the nacelle ............................................................183

CHAPTER 8 - CFD ANALYSIS OF OWC PNEUMATIC SYSTEM ..........................................214

8.1 PERFORMANCE ANALYSIS OF A COMPLETE OWC PNEUMATIC SYSTEM .....................................214

CHAPTER 9 - COMPARISON OF NUMERICAL ANALYSES AND FIELD DATA FROM A

CHAPTER 10 - CONCLUSIONS AND RECOMMENDATIONS .................................................249

10.1 CONCLUSIONS ...........................................................................................................................249

APPENDIX A-1 .........................................................................................................................................260

APPENDIX A-2 .........................................................................................................................................263

APPENDIX A-3 .........................................................................................................................................264

APPENDIX A-4 .........................................................................................................................................265

APPENDIX A-5 .........................................................................................................................................266

APPENDIX A-6 .........................................................................................................................................267

APPENDIX A-7 .........................................................................................................................................268

APPENDIX A-8 .........................................................................................................................................272

Figure 7.4 Distribution of streamlines in vicinity of a Wells turbine blade with γ = 3° at

Figure 8.13 Effect of nozzle-diffuser length on pneumatic efficiency, ηpneu of OWC

A = Planform area of the aerofoil section [m2]

c = Chord length of blade [m]

P* = Non-dimensional pressure drop across turbine [--]

U = Flow velocity [m/s]

αeq = Angle of equivalent incidence [deg]

αm = Mean angle of attack [deg]

αs = Stall angle [deg]

β = Angle of incidence at WR [deg]

δ0 = Cascade interference coefficient for Drag [--]

“” ____ 2011